Replication Notebook

Selective and Incentive Effects of Elections: Evidence from State Supreme Courts

Elliott Ash and Bentley MacLeod

This notebook includes the stata code for replicating the tables and figures in the text and appendix.


Set up packages and directories

In [161]:
*capture quietly ssc install fsum
*capture quietly ssc install ivreg2
*capture quietly ssc install ftools
*capture quietly ssc install reghdfe
*capture quietly ssc install cmogram
*capture quietly ssc install binscatter
*capture quietly ssc install estout
In [162]:
capture quietly mkdir 0-summ-stats
capture quietly mkdir 1-select

Load data and set macros

In [163]:
set more off
set matsize 10000
capture quietly restore

grstyle init
grstyle set plain, horizontal grid

clear all
use elections_ready, clear
keep if select_include

local yvars "action_affirm_m txt_words txt_words_m txt_vocab_m txt_vocabnorm_m txt_words_per_sent txt_chars_per_word dis_numops con_numops proc_pubdays_m toc_total_m toc_unique_m casetype_admin_m casetype_con_m casetype_crim_m casetype_civ_m"
local qvars "numops cites_pos cites_pos_m cites_neg cites_neg_m cites_outstate cites_outstate_m cites_quote cites_quote_m cites_discuss cites_discuss_m cites_overruled cites_overruled_m cites_supersede cites_supersede_m cites10_pos cites10_pos_m"
local rule_controls "rule_adminoffice rule_iac rule_judgesup rule_ntou rule_termdown rule_termup rule_utop rule_ptonp rule_judicialexpend"
local judge_controls "x_startage x_female x_democrat x_republican x_topschool x_startage_missing   x_topschool_missing"
local mainvars "txt_words_norm cites_pos_m_norm"
local additional_vars " txt_chars_norm txt_sents_norm  toc_total_norm cites_all_m_norm cites_discuss_m_norm  cites_outstate_m_norm"
local morevars "disc_numops dis_numops con_numops casetype_admin_m casetype_crim_m casetype_con_m casetype_civ_m txt_words_m toc_unique_m_norm  cites_disting_m cites_neg_m cites_circuit_m cites_multuse_m proc_pubdays_m"
local main_st "txt_words_norm_st cites_pos_m_norm_st" 
local main_j "txt_words_norm_j cites_pos_m_norm_j"
local additional_st "txt_chars_norm_st txt_sents_norm_st  cites_all_m_norm_st cites_discuss_m_norm_st  cites_outstate_m_norm_st" 
local additional_j "txt_chars_norm_j txt_sents_norm_j  cites_all_m_norm_j cites_discuss_m_norm_j cites_outstate_m_norm_j"







(6 observations deleted)











Summary Statistics

List variables with labels.

In [164]:
foreach var of varlist `qvars' `yvars' `rule_controls' `judge_controls' {
di "`var'" _col(20) "`: var l `var''" _col(50)  "`: val l `var''"
}
numops             number of opinions written    
cites_pos          positive citations to opinion 
cites_pos_m        positive citations to opinion (mean)
cites_neg          negative citations to opinion 
cites_neg_m        negative citations to opinion (mean)
cites_outstate     positive citations to opinion by courts in other states
cites_outstate_m   positive citations to opinion by courts in other states (mean
> )                                              
cites_quote        quoted-directly citations to opinion
cites_quote_m      quoted-directly citations to opinion (mean)
cites_discuss      discussed-in citations to opinion
cites_discuss_m    discussed-in citations to opinion (mean)
cites_overruled    number of future cases that overrule this opinion
cites_overruled_m  number of future cases that overrule this opinion (mean)
cites_supersede    number of times the ruling is superseded by statute
cites_supersede_m  number of times the ruling is superseded by statute (mean)
cites10_pos        positive citations (within ten years) to opinion
cites10_pos_m      positive citations (within ten years) to opinion (mean)
action_affirm_m    =1: affirms lower-court opinion (mean)
txt_words          number of words written in opinions
txt_words_m        number of words written in opinions (mean)
txt_vocab_m        number of unique words used (vocabulary size) (mean)
txt_vocabnorm_m    unique words used, normalized by length of opinion (in words)
>  (mean)                                        
txt_words_per_sent Sentence length -- words per sentence
txt_chars_per_word Word length -- characters per word
dis_numops         Number of dissents written    
con_numops         Number of concurrences written
proc_pubdays_m     number of days between submission/arguent and publication; ve
> ry sparse (mean)                               
toc_total_m        total previous cases cited (including multiples) (mean)
toc_unique_m       number of unique previous cases cited (length of table of cas
> es) (mean)                                     
casetype_admin_m   legal area is most related to administrative law (hand-coded 
> from legal topic) (                            
casetype_con_m     legal area is most related to constitutional law (hand-coded 
> from legal topic) (                            
casetype_crim_m    legal area is most related to criminal law (hand-coded from l
> egal topic) (mean)                             
casetype_civ_m     legal area is most related to civil law (hand-coded from lega
> l topic) (mean)                                
rule_adminoffice   =1 if there is a court administrative office
rule_iac           =1 if there is an intermediate appellate court operating
rule_judgesup      =1 after number of judges increased.
rule_ntou          =1 after move from no-retention to uncontested.
rule_termdown      =1 after term length goes down
rule_termup        =1 after term length goes up  
rule_utop          =1 after move from uncontested to partisan
rule_ptonp         =1 after move from partisan to nonpartisan
rule_judicialexpendlog state govt expenditures on judiciary
x_startage         judge starting age            
x_female           =1 if judge is female         
x_democrat         =1 if judge is democrat       
x_republican       =1 if judge is republican     
x_topschool        =1 if judge went to top law school
x_startage_missing =1 if start age is missing    
x_topschool_missing=1 if law school missing      

Summary statistics, for all judges.

In [165]:
fsum `qvars' `yvars' `rule_controls' `judge_controls'
            Variable |        N     Mean       SD      Min      Max                                                                                                                              
---------------------+---------------------------------------------
              numops |    15004    25.73    15.86     5.00   336.00  
           cites_pos |    15004   297.85   273.49     0.00  6133.00  
         cites_pos_m |    15004    13.06    12.52     0.00   342.54  
           cites_neg |    15004    11.96    15.89     0.00   246.00  
         cites_neg_m |    15004     0.58     0.88     0.00    20.62  
      cites_outstate |    15004    44.61    81.29     0.00  2308.00  
    cites_outstate_m |    15004     1.82     2.34     0.00    79.59  
         cites_quote |    15004    71.67    72.11     0.00  1525.00  
       cites_quote_m |    15004     3.28     3.73     0.00    86.91  
       cites_discuss |    15004    66.80    53.54     0.00   902.00  
     cites_discuss_m |    15004     2.94     2.37     0.00    53.06  
     cites_overruled |    15004     1.08     3.20     0.00   116.00  
   cites_overruled_m |    15004     0.05     0.18     0.00     5.52  
     cites_supersede |    15004     0.98     1.90     0.00    26.00  
   cites_supersede_m |    15004     0.05     0.10     0.00     1.20  
         cites10_pos |    15004   152.40   142.46     0.00  2242.00  
       cites10_pos_m |    15004     6.35     5.21     0.00    88.00  
     action_affirm_m |    15004     0.51     0.18     0.00     1.00  
           txt_words |    15004 56365.15 32537.94   298.00 3.45e+05  
         txt_words_m |    15004  2435.04  1225.66    59.60 15679.08  
         txt_vocab_m |    15004   682.22   236.12    44.40  2623.94  
     txt_vocabnorm_m |    15004     0.33     0.06     0.16     0.81  
  txt_words_per_sent |    15004    23.10     3.96    10.04    52.36  
  txt_chars_per_word |    15004     4.61     0.11     4.19     5.06  
          dis_numops |    15004     3.90     5.77     0.00   129.00  
          con_numops |    15004     2.50     4.40     0.00    65.00  
      proc_pubdays_m |    15004    21.04    47.22     0.00   457.00  
         toc_total_m |    15004    26.07    20.04     0.33   275.29  
        toc_unique_m |    15004    22.60    15.91     0.33   236.86  
    casetype_admin_m |    15004     0.08     0.08     0.00     0.67  
      casetype_con_m |    15004     0.10     0.09     0.00     0.80  
     casetype_crim_m |    15004     0.27     0.21     0.00     1.00  
      casetype_civ_m |    15004     0.51     0.20     0.00     1.00  
    rule_adminoffice |    15004     0.58     0.49     0.00     1.00  
            rule_iac |    15004     0.24     0.43     0.00     1.00  
       rule_judgesup |    15004     0.02     0.15     0.00     1.00  
           rule_ntou |    15004     0.01     0.11     0.00     1.00  
       rule_termdown |    15004     0.95     0.23     0.00     1.00  
         rule_termup |    15004     0.05     0.22     0.00     1.00  
           rule_utop |    15004     0.01     0.08     0.00     1.00  
          rule_ptonp |    15004     0.04     0.19     0.00     1.00  
 rule_judicialexpend |    13857     8.95     1.85     4.88    14.02  
          x_startage |    15004    47.47    16.41     0.00    86.00  
            x_female |    15004     0.03     0.17     0.00     1.00  
          x_democrat |    15004     0.33     0.47     0.00     1.00  
        x_republican |    15004     0.17     0.38     0.00     1.00  
         x_topschool |    15004     0.17     0.37     0.00     1.00  
  x_startage_missing |    15004     0.09     0.28     0.00     1.00  
 x_topschool_missing |    15004     0.18     0.38     0.00     1.00  

Summary statistics in partisan election systems.

In [166]:
fsum `qvars' `yvars' `rule_controls' `judge_controls' if rule_howselect == "P"
            Variable |        N     Mean       SD      Min      Max                                                                                                                              
---------------------+---------------------------------------------
              numops |     6271    28.85    18.22     5.00   160.00  
           cites_pos |     6271   320.15   319.11     0.00  4821.00  
         cites_pos_m |     6271    13.17    13.97     0.00   341.55  
           cites_neg |     6271    12.93    16.43     0.00   170.00  
         cites_neg_m |     6271     0.63     0.98     0.00    20.62  
      cites_outstate |     6271    56.75   119.90     0.00  2308.00  
    cites_outstate_m |     6271     2.00     3.11     0.00    79.59  
         cites_quote |     6271    69.96    78.69     0.00  1525.00  
       cites_quote_m |     6271     2.99     3.70     0.00    86.91  
       cites_discuss |     6271    65.92    58.86     0.00   857.00  
     cites_discuss_m |     6271     2.72     2.35     0.00    41.11  
     cites_overruled |     6271     1.25     3.55     0.00   116.00  
   cites_overruled_m |     6271     0.06     0.19     0.00     5.52  
     cites_supersede |     6271     0.82     1.41     0.00    14.00  
   cites_supersede_m |     6271     0.04     0.07     0.00     0.83  
         cites10_pos |     6271   157.38   164.24     0.00  2242.00  
       cites10_pos_m |     6271     6.07     5.56     0.00    71.47  
     action_affirm_m |     6271     0.52     0.17     0.00     1.00  
           txt_words |     6271 54931.18 31475.11   298.00 2.41e+05  
         txt_words_m |     6271  2152.15  1018.42    59.60  9166.27  
         txt_vocab_m |     6271   615.34   201.78    44.40  1653.11  
     txt_vocabnorm_m |     6271     0.35     0.06     0.18     0.81  
  txt_words_per_sent |     6271    24.01     4.40    10.28    52.36  
  txt_chars_per_word |     6271     4.59     0.10     4.19     4.96  
          dis_numops |     6271     4.88     7.13     0.00   129.00  
          con_numops |     6271     2.74     4.87     0.00    65.00  
      proc_pubdays_m |     6271    22.05    54.51     0.00   447.19  
         toc_total_m |     6271    21.90    15.60     0.33   150.38  
        toc_unique_m |     6271    19.11    12.81     0.33   125.00  
    casetype_admin_m |     6271     0.08     0.08     0.00     0.60  
      casetype_con_m |     6271     0.10     0.08     0.00     0.80  
     casetype_crim_m |     6271     0.27     0.24     0.00     1.00  
      casetype_civ_m |     6271     0.51     0.21     0.00     1.00  
    rule_adminoffice |     6271     0.47     0.50     0.00     1.00  
            rule_iac |     6271     0.16     0.37     0.00     1.00  
       rule_judgesup |     6271     0.00     0.00     0.00     0.00  
           rule_ntou |     6271     0.03     0.17     0.00     1.00  
       rule_termdown |     6271     1.00     0.00     1.00     1.00  
         rule_termup |     6271     0.03     0.18     0.00     1.00  
           rule_utop |     6271     0.01     0.12     0.00     1.00  
          rule_ptonp |     6271     0.00     0.00     0.00     0.00  
 rule_judicialexpend |     5647     8.84     1.72     5.73    12.60  
          x_startage |     6271    47.70    16.60     0.00    82.00  
            x_female |     6271     0.02     0.13     0.00     1.00  
          x_democrat |     6271     0.52     0.50     0.00     1.00  
        x_republican |     6271     0.17     0.37     0.00     1.00  
         x_topschool |     6271     0.12     0.33     0.00     1.00  
  x_startage_missing |     6271     0.09     0.28     0.00     1.00  
 x_topschool_missing |     6271     0.23     0.42     0.00     1.00  

Summary statistic in non-partisan election systems.

In [167]:
fsum `qvars' `yvars' `rule_controls' `judge_controls' if rule_howselect == "NP"
            Variable |        N     Mean       SD      Min      Max                                                                                                                              
---------------------+---------------------------------------------
              numops |     3372    24.34    12.82     5.00   157.00  
           cites_pos |     3372   234.16   161.54     0.00  2197.00  
         cites_pos_m |     3372    10.23     5.97     0.00    69.33  
           cites_neg |     3372     8.46     9.78     0.00    82.00  
         cites_neg_m |     3372     0.39     0.52     0.00     6.86  
      cites_outstate |     3372    33.03    22.75     0.00   178.00  
    cites_outstate_m |     3372     1.44     0.85     0.00     6.86  
         cites_quote |     3372    56.63    45.61     0.00   696.00  
       cites_quote_m |     3372     2.52     1.99     0.00    30.00  
       cites_discuss |     3372    62.01    44.31     0.00   470.00  
     cites_discuss_m |     3372     2.72     1.69     0.00    22.71  
     cites_overruled |     3372     1.03     2.95     0.00    62.00  
   cites_overruled_m |     3372     0.05     0.18     0.00     3.44  
     cites_supersede |     3372     0.75     1.32     0.00    13.00  
   cites_supersede_m |     3372     0.03     0.07     0.00     0.80  
         cites10_pos |     3372   118.18    96.11     0.00  1509.00  
       cites10_pos_m |     3372     4.93     3.10     0.00    33.57  
     action_affirm_m |     3372     0.57     0.15     0.00     1.00  
           txt_words |     3372 56464.78 31817.04  2938.00 2.31e+05  
         txt_words_m |     3372  2491.45  1120.33   449.90  9776.67  
         txt_vocab_m |     3372   691.93   202.41   219.87  1923.78  
     txt_vocabnorm_m |     3372     0.32     0.05     0.16     0.55  
  txt_words_per_sent |     3372    23.06     3.28    10.04    38.71  
  txt_chars_per_word |     3372     4.60     0.12     4.21     5.06  
          dis_numops |     3372     3.67     4.63     0.00    52.00  
          con_numops |     3372     2.43     3.64     0.00    47.00  
      proc_pubdays_m |     3372    17.50    37.98     0.00   457.00  
         toc_total_m |     3372    24.69    14.27     0.93   225.78  
        toc_unique_m |     3372    21.78    11.61     0.87   172.33  
    casetype_admin_m |     3372     0.08     0.08     0.00     0.67  
      casetype_con_m |     3372     0.10     0.08     0.00     0.60  
     casetype_crim_m |     3372     0.23     0.16     0.00     1.00  
      casetype_civ_m |     3372     0.57     0.17     0.00     1.00  
    rule_adminoffice |     3372     0.52     0.50     0.00     1.00  
            rule_iac |     3372     0.28     0.45     0.00     1.00  
       rule_judgesup |     3372     0.07     0.25     0.00     1.00  
           rule_ntou |     3372     0.00     0.00     0.00     0.00  
       rule_termdown |     3372     1.00     0.00     1.00     1.00  
         rule_termup |     3372     0.05     0.22     0.00     1.00  
           rule_utop |     3372     0.00     0.00     0.00     0.00  
          rule_ptonp |     3372     0.12     0.33     0.00     1.00  
 rule_judicialexpend |     3101     8.13     1.78     4.88    12.06  
          x_startage |     3372    48.96    14.57     0.00    77.00  
            x_female |     3372     0.03     0.17     0.00     1.00  
          x_democrat |     3372     0.19     0.39     0.00     1.00  
        x_republican |     3372     0.19     0.39     0.00     1.00  
         x_topschool |     3372     0.12     0.33     0.00     1.00  
  x_startage_missing |     3372     0.06     0.24     0.00     1.00  
 x_topschool_missing |     3372     0.12     0.33     0.00     1.00  

Summary statistics in merit systems.

In [168]:
fsum `qvars' `yvars' `rule_controls' `judge_controls' if rule_howselect == "M"
            Variable |        N     Mean       SD      Min      Max                                                                                                                              
---------------------+---------------------------------------------
              numops |     2442    25.20    17.12     5.00   336.00  
           cites_pos |     2442   345.96   231.20     0.00  5297.00  
         cites_pos_m |     2442    15.08    10.34     0.00   311.59  
           cites_neg |     2442    18.21    21.93     0.00   246.00  
         cites_neg_m |     2442     0.84     1.11     0.00    10.00  
      cites_outstate |     2442    44.06    37.44     0.00   986.00  
    cites_outstate_m |     2442     2.04     2.04     0.00    58.00  
         cites_quote |     2442    86.84    63.03     0.00   633.00  
       cites_quote_m |     2442     3.99     3.27     0.00    39.56  
       cites_discuss |     2442    84.13    53.52     0.00   902.00  
     cites_discuss_m |     2442     3.69     2.27     0.00    53.06  
     cites_overruled |     2442     1.38     3.57     0.00    48.00  
   cites_overruled_m |     2442     0.07     0.22     0.00     4.00  
     cites_supersede |     2442     2.08     3.35     0.00    26.00  
   cites_supersede_m |     2442     0.10     0.17     0.00     1.20  
         cites10_pos |     2442   204.72   137.16     0.00  1496.00  
       cites10_pos_m |     2442     8.62     4.83     0.00    88.00  
     action_affirm_m |     2442     0.50     0.18     0.00     1.00  
           txt_words |     2442 67706.67 37062.25  4115.00 3.45e+05  
         txt_words_m |     2442  2987.94  1568.26   610.34 15679.08  
         txt_vocab_m |     2442   802.91   281.18   236.62  2623.94  
     txt_vocabnorm_m |     2442     0.32     0.05     0.19     0.55  
  txt_words_per_sent |     2442    21.42     2.69    14.40    42.60  
  txt_chars_per_word |     2442     4.67     0.10     4.29     5.03  
          dis_numops |     2442     4.36     5.18     0.00    47.00  
          con_numops |     2442     3.89     5.58     0.00    56.00  
      proc_pubdays_m |     2442     3.68    21.82     0.00   229.31  
         toc_total_m |     2442    36.12    27.81     3.21   275.29  
        toc_unique_m |     2442    31.15    22.09     3.08   236.86  
    casetype_admin_m |     2442     0.08     0.08     0.00     0.55  
      casetype_con_m |     2442     0.13     0.10     0.00     0.67  
     casetype_crim_m |     2442     0.37     0.24     0.00     1.00  
      casetype_civ_m |     2442     0.42     0.20     0.00     1.00  
    rule_adminoffice |     2442     0.89     0.31     0.00     1.00  
            rule_iac |     2442     0.51     0.50     0.00     1.00  
       rule_judgesup |     2442     0.05     0.22     0.00     1.00  
           rule_ntou |     2442     0.00     0.00     0.00     0.00  
       rule_termdown |     2442     1.00     0.00     1.00     1.00  
         rule_termup |     2442     0.08     0.27     0.00     1.00  
           rule_utop |     2442     0.00     0.00     0.00     0.00  
          rule_ptonp |     2442     0.07     0.26     0.00     1.00  
 rule_judicialexpend |     2386     9.98     1.43     6.98    13.89  
          x_startage |     2442    48.18    11.42     0.00    71.00  
            x_female |     2442     0.05     0.21     0.00     1.00  
          x_democrat |     2442     0.29     0.46     0.00     1.00  
        x_republican |     2442     0.15     0.36     0.00     1.00  
         x_topschool |     2442     0.14     0.35     0.00     1.00  
  x_startage_missing |     2442     0.03     0.17     0.00     1.00  
 x_topschool_missing |     2442     0.06     0.23     0.00     1.00  

Tabulation of judge-year observations by judge selection system. G = governor, L = legislature, M = merit, NP = nonpartisan, and P = partisan.

In [169]:
tab xc_howselected
xc_howselec |
        ted |      Freq.     Percent        Cum.
------------+-----------------------------------
          G |      1,877       15.11       15.11
          L |        511        4.11       19.23
          M |      1,593       12.83       32.05
         NP |      2,991       24.08       56.14
          P |      5,448       43.86      100.00
------------+-----------------------------------
      Total |     12,420      100.00

Tabulation of judge-year observations by judge retention system. G=governor, L=legislature, T=life tenure, P = partisan elections, NP = nonpartisan elections, U = uncontested elections.


Output and Quality

Compute judge fixed effects from residuals, and plot histograms for work output and work quality.

In [170]:
qui areg txt_words, absorb(stateyearfe)
predict resids_output, residuals
bys judgenum: egen resids_judgemean_output = mean(resids_output)
In [171]:
hist resids_judgemean_output, xtitle("Judge Fixed Effects: Total Words Written")
graph export 0-summ-stats/hist-resids-output.pdf, replace
(bin=41, start=-71956.125, width=4254.9615)

(file 0-summ-stats/hist-resids-output.pdf written in PDF format)
In [172]:
qui areg cites_pos_m, absorb(stateyearfe)
predict resids_quality, residuals
bys judgenum: egen resids_judgemean_quality = mean(resids_quality)
In [173]:
hist resids_judgemean_quality, xtitle("Judge Fixed Effects: Positive Cites Per Case")
graph export 0-summ-stats/hist-resids-quality.pdf, replace
(bin=41, start=-37.319912, width=5.4625414)

(file 0-summ-stats/hist-resids-quality.pdf written in PDF format)
In [174]:
binscatter pct_cites_pos_m pct_txt_words, xtitle("Work Output (Rank Percentile)") ytitle("Work Quality (Rank Percentile)") xsize(7) 
graph export 0-summ-stats/quality-vs-output-pct.pdf, replace
graph display
warning: nquantiles(20) was specified, but only 13 were generated. see help file
>  under nquantiles() for explanation.

(file 0-summ-stats/quality-vs-output-pct.pdf written in PDF format)


Bar association evaluations

In [175]:
su x_rating_good if year >= 1988
    Variable |        Obs        Mean    Std. Dev.       Min        Max
-------------+---------------------------------------------------------
x_rating_g~d |        170    .8705882    .3366467          0          1

Plot work quality and work output versus bar evaluations.

In [176]:
bysort state: egen rating_m = mean(x_rating_avg)
bysort state: egen rating_sd = sd(x_rating_avg)
gen ratingnorm = (x_rating_avg - rating_m) / rating_sd
(11835 missing values generated)

(11835 missing values generated)

(14,768 missing values generated)
In [177]:
binscatter pct_cites_pos_m x_rating_good if year >= 1988, absorb(stateyear)  xtitle("Normalized ABA Rating") ytitle("Work Quality") 
graph export 0-summ-stats/binscatter-bar-quality.pdf, replace
graph display

(file 0-summ-stats/binscatter-bar-quality.pdf written in PDF format)

In [178]:
binscatter pct_txt_words x_rating_good if year >= 1988, absorb(stateyear)  xtitle("Normalized ABA Rating") ytitle("Work Output") 
graph export 0-summ-stats/binscatter-bar-output.pdf, replace
graph display

(file 0-summ-stats/binscatter-bar-output.pdf written in PDF format)

Number of "Treatment" and "Control" Judges

In [179]:
bys stateyear:  egen ptomselect_avg = mean(ptomselect)
bys stateyear:  egen ptonpselect_avg = mean(ptonpselect)
bys stateyear:  egen nptomselect_avg = mean(nptomselect)

gen treat_year_ptom = ptomselect_avg > 0 & ptomselect_avg < 1
gen treat_year_ptonp = ptonpselect_avg > 0 & ptonpselect_avg < 1
gen treat_year_nptom = nptomselect_avg > 0 & nptomselect_avg < 1
In [180]:
distinct stateyear if treat_year_ptom
distinct stateyear if treat_year_ptonp
distinct stateyear if treat_year_nptom

             |        Observations
             |      total   distinct
-------------+----------------------
 stateyearfe |       1051        164


             |        Observations
             |      total   distinct
-------------+----------------------
 stateyearfe |        195         33


             |        Observations
             |      total   distinct
-------------+----------------------
 stateyearfe |        354         63
In [181]:
distinct judgenum if treat_year_ptom & !ptomselect
distinct judgenum if treat_year_ptom & ptomselect
distinct judgenum if treat_year_ptom

          |        Observations
          |      total   distinct
----------+----------------------
 judgenum |        423         45


          |        Observations
          |      total   distinct
----------+----------------------
 judgenum |        628         67


          |        Observations
          |      total   distinct
----------+----------------------
 judgenum |       1051        112
In [182]:
distinct judgenum if treat_year_ptonp & !ptonpselect
distinct judgenum if treat_year_ptonp & ptonpselect
distinct judgenum if treat_year_ptonp

          |        Observations
          |      total   distinct
----------+----------------------
 judgenum |         91         17


          |        Observations
          |      total   distinct
----------+----------------------
 judgenum |        104         21


          |        Observations
          |      total   distinct
----------+----------------------
 judgenum |        195         38
In [183]:
distinct judgenum if treat_year_nptom & !nptomselect
distinct judgenum if treat_year_nptom & nptomselect
distinct judgenum if treat_year_nptom

          |        Observations
          |      total   distinct
----------+----------------------
 judgenum |        154         19


          |        Observations
          |      total   distinct
----------+----------------------
 judgenum |        200         27


          |        Observations
          |      total   distinct
----------+----------------------
 judgenum |        354         46

Results

Main Regression Table

In [184]:
eststo clear

* baseline
eststo: reghdfe pct_cites_pos_m  ptomselect ptonpselect nptomselect  , absorb(stateyear) cluster(state year)

* + cohort FEs and trends (including birth)
eststo: reghdfe pct_cites_pos_m  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) ) cluster(state year)

* + experience controls

eststo: reghdfe pct_cites_pos_m  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) xt_exper) cluster(state year)

(MWFE estimator converged in 1 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.

HDFE Linear regression                            Number of obs   =     14,996
Absorbing 1 HDFE group                            F(   3,     47) =       7.80
Statistics robust to heteroskedasticity           Prob > F        =     0.0003
                                                  R-squared       =     0.0040
                                                  Adj R-squared   =    -0.1896
Number of clusters (state)   =         50         Within R-sq.    =     0.0039
Number of clusters (year)    =         48         Root MSE        =     0.3722

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_~s_pos_m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .1776986   .0429922     4.13   0.000     .0912094    .2641878
 ptonpselect |   .1010445   .0476802     2.12   0.039     .0051242    .1969647
 nptomselect |  -.0686153   .1364205    -0.50   0.617    -.3430579    .2058273
       _cons |   .4878864   .0033343   146.32   0.000     .4811787    .4945941
------------------------------------------------------------------------------

Absorbed degrees of freedom:
-----------------------------------------------------+
 Absorbed FE | Categories  - Redundant  = Num. Coefs |
-------------+---------------------------------------|
 stateyearfe |      2438        2438           0    *|
-----------------------------------------------------+
* = FE nested within cluster; treated as redundant for DoF computation
(est1 stored)

(dropped 10 singleton observations)
(MWFE estimator converged in 27 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.

HDFE Linear regression                            Number of obs   =     14,894
Absorbing 4 HDFE groups                           F(   3,     47) =       3.12
Statistics robust to heteroskedasticity           Prob > F        =     0.0346
                                                  R-squared       =     0.0456
                                                  Adj R-squared   =    -0.1517
Number of clusters (state)   =         50         Within R-sq.    =     0.0012
Number of clusters (year)    =         48         Root MSE        =     0.3662

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_~s_pos_m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .0982243   .0401769     2.44   0.018     .0173987    .1790499
 ptonpselect |   .0956246   .0529564     1.81   0.077      -.01091    .2021592
 nptomselect |  -.0886809   .1534732    -0.58   0.566    -.3974291    .2200672
       _cons |   .4930914   .0035484   138.96   0.000     .4859529      .50023
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est2 stored)

(dropped 14 singleton observations)
(MWFE estimator converged in 31 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.

HDFE Linear regression                            Number of obs   =     14,890
Absorbing 5 HDFE groups                           F(   3,     47) =       2.59
Statistics robust to heteroskedasticity           Prob > F        =     0.0639
                                                  R-squared       =     0.0512
                                                  Adj R-squared   =    -0.1481
Number of clusters (state)   =         50         Within R-sq.    =     0.0011
Number of clusters (year)    =         48         Root MSE        =     0.3656

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_~s_pos_m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |    .083149   .0408577     2.04   0.048     .0009539    .1653441
 ptonpselect |   .0866382   .0472728     1.83   0.073    -.0084624    .1817389
 nptomselect |   -.106856   .1553657    -0.69   0.495    -.4194115    .2056995
       _cons |   .4945942   .0036461   135.65   0.000     .4872592    .5019292
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
                 xt_exper |        34           1          33    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est3 stored)
In [185]:
%html
esttab, r2 se star(+ 0.10 * 0.05 ** 0.01) drop(   _cons) html
esttab using "1-select/1a-main-regs-quality-stars.csv", replace r2 se star(+ 0.10 * 0.05 ** 0.01) drop(  _cons)

(1) (2) (3)
pct_cites_pos_m pct_cites_pos_m pct_cites_pos_m

ptomselect 0.178** 0.0982* 0.0831*
(0.0430) (0.0402) (0.0409)
 
ptonpselect 0.101* 0.0956+ 0.0866+
(0.0477) (0.0530) (0.0473)
 
nptomselect -0.0686 -0.0887 -0.107
(0.136) (0.153) (0.155)

N 14996 14894 14890
R2 0.004 0.046 0.051

Standard errors in parentheses
+ p < 0.10, * p < 0.05, ** p < 0.01
(output written to 1-select/1a-main-regs-quality-stars.csv)
In [186]:
eststo clear

* baseline
eststo: reghdfe pct_txt_words  ptomselect ptonpselect nptomselect  , absorb(stateyear) cluster(state year)

* + cohort FEs and trends (including birth)
eststo: reghdfe pct_txt_words  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) ) cluster(state year)

* + experience controls

eststo: reghdfe pct_txt_words  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) xt_exper) cluster(state year)

* +state case controls
* eststo: reghdfe pct_txt_words  ptomselect ptonpselect nptomselect  , absorb(stateyear  x_decade_start birth_decade statefe##c.(x_decade_start casetype_*_m pca_* pct_numops) xt_exper ) cluster(state year)

(MWFE estimator converged in 1 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.
warning: missing F statistic; dropped variables due to collinearity or too few c
> lusters

HDFE Linear regression                            Number of obs   =     14,996
Absorbing 1 HDFE group                            F(   3,     47) =          .
Statistics robust to heteroskedasticity           Prob > F        =          .
                                                  R-squared       =     0.0008
                                                  Adj R-squared   =    -0.1934
Number of clusters (state)   =         50         Within R-sq.    =     0.0008
Number of clusters (year)    =         48         Root MSE        =     0.3730

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_txt_wo~s |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .0631543   .0348421     1.81   0.076     -.006939    .1332477
 ptonpselect |   .0886494   .0139064     6.37   0.000     .0606734    .1166255
 nptomselect |  -.0628946   .0732454    -0.86   0.395    -.2102454    .0844562
       _cons |   .4951858   .0022055   224.52   0.000     .4907488    .4996227
------------------------------------------------------------------------------

Absorbed degrees of freedom:
-----------------------------------------------------+
 Absorbed FE | Categories  - Redundant  = Num. Coefs |
-------------+---------------------------------------|
 stateyearfe |      2438        2438           0    *|
-----------------------------------------------------+
* = FE nested within cluster; treated as redundant for DoF computation
(est1 stored)

(dropped 10 singleton observations)
(MWFE estimator converged in 27 iterations)

HDFE Linear regression                            Number of obs   =     14,894
Absorbing 4 HDFE groups                           F(   3,     47) =       2.47
Statistics robust to heteroskedasticity           Prob > F        =     0.0734
                                                  R-squared       =     0.0240
                                                  Adj R-squared   =    -0.1778
Number of clusters (state)   =         50         Within R-sq.    =     0.0003
Number of clusters (year)    =         48         Root MSE        =     0.3703

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_txt_wo~s |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .0379152   .0393591     0.96   0.340    -.0412652    .1170955
 ptonpselect |   .0812301   .0463391     1.75   0.086    -.0119921    .1744523
 nptomselect |  -.0602099   .1275312    -0.47   0.639    -.3167695    .1963497
       _cons |   .4968775   .0029199   170.17   0.000     .4910035    .5027515
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est2 stored)

(dropped 14 singleton observations)
(MWFE estimator converged in 31 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.

HDFE Linear regression                            Number of obs   =     14,890
Absorbing 5 HDFE groups                           F(   3,     47) =       2.51
Statistics robust to heteroskedasticity           Prob > F        =     0.0704
                                                  R-squared       =     0.0585
                                                  Adj R-squared   =    -0.1393
Number of clusters (state)   =         50         Within R-sq.    =     0.0004
Number of clusters (year)    =         48         Root MSE        =     0.3642

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_txt_wo~s |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .0279854   .0421183     0.66   0.510    -.0567458    .1127165
 ptonpselect |   .1000869   .0476567     2.10   0.041      .004214    .1959599
 nptomselect |  -.0717419   .1308154    -0.55   0.586    -.3349086    .1914248
       _cons |   .4972489   .0031275   158.99   0.000     .4909572    .5035405
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
                 xt_exper |        34           1          33    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est3 stored)
In [187]:
%html
esttab, r2 se star(+ 0.10 * 0.05 ** 0.01) drop(   _cons) html
esttab using "1-select/1a-main-regs-output-stars.csv", replace r2 se star(+ 0.10 * 0.05 ** 0.01) drop(  _cons)

(1) (2) (3)
pct_txt_words pct_txt_words pct_txt_words

ptomselect 0.0632+ 0.0379 0.0280
(0.0348) (0.0394) (0.0421)
 
ptonpselect 0.0886** 0.0812+ 0.100*
(0.0139) (0.0463) (0.0477)
 
nptomselect -0.0629 -0.0602 -0.0717
(0.0732) (0.128) (0.131)

N 14996 14894 14890
R2 0.001 0.024 0.058

Standard errors in parentheses
+ p < 0.10, * p < 0.05, ** p < 0.01
(output written to 1-select/1a-main-regs-output-stars.csv)
In [188]:
%html
esttab, r2 se star(+ 0.10 * 0.05 ** 0.01) drop(   _cons) html

(1) (2) (3)
pct_txt_words pct_txt_words pct_txt_words

ptomselect 0.0632+ 0.0379 0.0280
(0.0348) (0.0394) (0.0421)
 
ptonpselect 0.0886** 0.0812+ 0.100*
(0.0139) (0.0463) (0.0477)
 
nptomselect -0.0629 -0.0602 -0.0717
(0.0732) (0.128) (0.131)

N 14996 14894 14890
R2 0.001 0.024 0.058

Standard errors in parentheses
+ p < 0.10, * p < 0.05, ** p < 0.01
In [189]:
eststo clear
eststo: reghdfe pct_cites_pos_m  ptomselect ptonpselect nptomselect  , absorb(stateyear) cluster(state year)  
eststo: reghdfe pct_cites_pos_m  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start) cluster(state year)  
eststo: reghdfe pct_cites_pos_m  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start statefe##c.x_decade_start) cluster(state year)  
eststo: reghdfe pct_cites_pos_m  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start statefe##c.x_decade_start one##c.(pca_*)) cluster(state year)

(MWFE estimator converged in 1 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.

HDFE Linear regression                            Number of obs   =     14,996
Absorbing 1 HDFE group                            F(   3,     47) =       7.80
Statistics robust to heteroskedasticity           Prob > F        =     0.0003
                                                  R-squared       =     0.0040
                                                  Adj R-squared   =    -0.1896
Number of clusters (state)   =         50         Within R-sq.    =     0.0039
Number of clusters (year)    =         48         Root MSE        =     0.3722

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_~s_pos_m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .1776986   .0429922     4.13   0.000     .0912094    .2641878
 ptonpselect |   .1010445   .0476802     2.12   0.039     .0051242    .1969647
 nptomselect |  -.0686153   .1364205    -0.50   0.617    -.3430579    .2058273
       _cons |   .4878864   .0033343   146.32   0.000     .4811787    .4945941
------------------------------------------------------------------------------

Absorbed degrees of freedom:
-----------------------------------------------------+
 Absorbed FE | Categories  - Redundant  = Num. Coefs |
-------------+---------------------------------------|
 stateyearfe |      2438        2438           0    *|
-----------------------------------------------------+
* = FE nested within cluster; treated as redundant for DoF computation
(est1 stored)

(MWFE estimator converged in 10 iterations)

HDFE Linear regression                            Number of obs   =     14,996
Absorbing 2 HDFE groups                           F(   3,     47) =       2.26
Statistics robust to heteroskedasticity           Prob > F        =     0.0936
                                                  R-squared       =     0.0188
                                                  Adj R-squared   =    -0.1727
Number of clusters (state)   =         50         Within R-sq.    =     0.0017
Number of clusters (year)    =         48         Root MSE        =     0.3696

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_~s_pos_m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .1061793   .0448004     2.37   0.022     .0160525    .1963062
 ptonpselect |   .0303277    .065369     0.46   0.645    -.1011777    .1618332
 nptomselect |  -.1099446   .1373455    -0.80   0.427    -.3862481    .1663588
       _cons |   .4943746      .0037   133.62   0.000     .4869311     .501818
------------------------------------------------------------------------------

Absorbed degrees of freedom:
--------------------------------------------------------+
    Absorbed FE | Categories  - Redundant  = Num. Coefs |
----------------+---------------------------------------|
    stateyearfe |      2438        2438           0    *|
 x_decade_start |         9           0           9     |
--------------------------------------------------------+
* = FE nested within cluster; treated as redundant for DoF computation
(est2 stored)

(MWFE estimator converged in 21 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.

HDFE Linear regression                            Number of obs   =     14,996
Absorbing 3 HDFE groups                           F(   3,     47) =       4.92
Statistics robust to heteroskedasticity           Prob > F        =     0.0047
                                                  R-squared       =     0.0334
                                                  Adj R-squared   =    -0.1650
Number of clusters (state)   =         50         Within R-sq.    =     0.0012
Number of clusters (year)    =         48         Root MSE        =     0.3683

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_~s_pos_m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .0846971   .0451077     1.88   0.067    -.0060478     .175442
 ptonpselect |   .1445027   .0422382     3.42   0.001     .0595304     .229475
 nptomselect |  -.0922859   .1605274    -0.57   0.568    -.4152254    .2306536
       _cons |   .4932936   .0037836   130.38   0.000     .4856819    .5009053
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2438        2438           0    *|
           x_decade_start |         9           0           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est3 stored)

(MWFE estimator converged in 25 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.

HDFE Linear regression                            Number of obs   =     14,996
Absorbing 4 HDFE groups                           F(   3,     47) =       4.00
Statistics robust to heteroskedasticity           Prob > F        =     0.0129
                                                  R-squared       =     0.0410
                                                  Adj R-squared   =    -0.1563
Number of clusters (state)   =         50         Within R-sq.    =     0.0011
Number of clusters (year)    =         48         Root MSE        =     0.3670

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_~s_pos_m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .0854482   .0441246     1.94   0.059     -.003319    .1742154
 ptonpselect |    .130514   .0432177     3.02   0.004     .0435712    .2174568
 nptomselect |  -.0828371   .1602857    -0.52   0.608    -.4052903    .2396162
       _cons |   .4933797   .0037543   131.42   0.000      .485827    .5009324
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2438        2438           0    *|
           x_decade_start |         9           0           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
                      one |         1           1           0     |
              one#c.pca_1 |         1           0           1    ?|
              one#c.pca_2 |         1           0           1    ?|
              one#c.pca_3 |         1           0           1    ?|
              one#c.pca_4 |         1           0           1    ?|
              one#c.pca_5 |         1           0           1    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est4 stored)
In [190]:
esttab using "1-select/1b-main-regs-quality.csv", replace r2 nostar cells(b(fmt(a3) star) se(par(( )) fmt(a2)) p(par([ ]) fmt(a2)) .) drop(   _cons)
esttab using "1-select/1b-main-regs-quality-stars.csv", replace r2 se star(+ 0.10 * 0.05 ** 0.01) drop(   _cons)
(note: file 1-select/1b-main-regs-quality.csv not found)
(output written to 1-select/1b-main-regs-quality.csv)

(note: file 1-select/1b-main-regs-quality-stars.csv not found)
(output written to 1-select/1b-main-regs-quality-stars.csv)
In [191]:
%html
esttab, r2 se star(+ 0.10 * 0.05 ** 0.01) drop(   _cons) html

(1) (2) (3) (4)
pct_cites_pos_m pct_cites_pos_m pct_cites_pos_m pct_cites_pos_m

ptomselect 0.178** 0.106* 0.0847+ 0.0854+
(0.0430) (0.0448) (0.0451) (0.0441)
 
ptonpselect 0.101* 0.0303 0.145** 0.131**
(0.0477) (0.0654) (0.0422) (0.0432)
 
nptomselect -0.0686 -0.110 -0.0923 -0.0828
(0.136) (0.137) (0.161) (0.160)

N 14996 14996 14996 14996
R2 0.004 0.019 0.033 0.041

Standard errors in parentheses
+ p < 0.10, * p < 0.05, ** p < 0.01

Additional Quality Variables

In [192]:
* cohort FE / trends spec
eststo clear

eststo: reghdfe cites_pos_m  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) ) cluster(state year)
eststo: reghdfe cites_pos_m_log  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) ) cluster(state year)
eststo: reghdfe pct_cites10_pos_m  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) ) cluster(state year)

eststo: reghdfe pct_cites_all_m  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) ) cluster(state year)
eststo: reghdfe pct_cites_discuss_m  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) ) cluster(state year)
eststo: reghdfe pct_cites_quote_m  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) ) cluster(state year)
eststo: reghdfe pct_cites_outstate_m  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) ) cluster(state year)

(dropped 16 singleton observations)
(MWFE estimator converged in 27 iterations)

HDFE Linear regression                            Number of obs   =     14,894
Absorbing 4 HDFE groups                           F(   3,     47) =       1.33
Statistics robust to heteroskedasticity           Prob > F        =     0.2755
                                                  R-squared       =     0.6179
                                                  Adj R-squared   =     0.5389
Number of clusters (state)   =         50         Within R-sq.    =     0.0002
Number of clusters (year)    =         48         Root MSE        =     8.5203

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
 cites_pos_m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .8868072   .4994771     1.78   0.082    -.1180111    1.891625
 ptonpselect |   .6496282   1.017616     0.64   0.526    -1.397552    2.696808
 nptomselect |  -.6503099   1.910484    -0.34   0.735    -4.493707    3.193087
       _cons |   13.05388   .0481985   270.84   0.000     12.95692    13.15085
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est1 stored)

(dropped 16 singleton observations)
(MWFE estimator converged in 27 iterations)

HDFE Linear regression                            Number of obs   =     14,894
Absorbing 4 HDFE groups                           F(   3,     47) =       2.02
Statistics robust to heteroskedasticity           Prob > F        =     0.1241
                                                  R-squared       =     0.8067
                                                  Adj R-squared   =     0.7667
Number of clusters (state)   =         50         Within R-sq.    =     0.0007
Number of clusters (year)    =         48         Root MSE        =     0.3207

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
~s_pos_m_log |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .0653707   .0371127     1.76   0.085    -.0092905    .1400318
 ptonpselect |   .0813504   .0451577     1.80   0.078    -.0094951    .1721959
 nptomselect |  -.0519415   .1014533    -0.51   0.611    -.2560391    .1521562
       _cons |   2.420225   .0030075   804.74   0.000     2.414174    2.426275
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est2 stored)

(dropped 10 singleton observations)
(MWFE estimator converged in 27 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.

HDFE Linear regression                            Number of obs   =     14,894
Absorbing 4 HDFE groups                           F(   3,     47) =       1.76
Statistics robust to heteroskedasticity           Prob > F        =     0.1668
                                                  R-squared       =     0.0434
                                                  Adj R-squared   =    -0.1545
Number of clusters (state)   =         50         Within R-sq.    =     0.0009
Number of clusters (year)    =         48         Root MSE        =     0.3666

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_~0_pos_m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |    .076682   .0443943     1.73   0.091    -.0126279    .1659919
 ptonpselect |   .0574457   .0423553     1.36   0.181    -.0277622    .1426535
 nptomselect |  -.1027733   .1147106    -0.90   0.375    -.3335413    .1279947
       _cons |   .4956407    .003267   151.71   0.000     .4890683    .5022131
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est3 stored)

(dropped 10 singleton observations)
(MWFE estimator converged in 27 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.

HDFE Linear regression                            Number of obs   =     14,894
Absorbing 4 HDFE groups                           F(   3,     47) =       2.86
Statistics robust to heteroskedasticity           Prob > F        =     0.0465
                                                  R-squared       =     0.0451
                                                  Adj R-squared   =    -0.1524
Number of clusters (state)   =         50         Within R-sq.    =     0.0010
Number of clusters (year)    =         48         Root MSE        =     0.3663

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_cite~l_m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .0943132   .0356336     2.65   0.011     .0226276    .1659987
 ptonpselect |   .0571979   .0536666     1.07   0.292    -.0507654    .1651613
 nptomselect |   -.083763   .1412771    -0.59   0.556    -.3679758    .2004498
       _cons |   .4938676   .0033143   149.01   0.000     .4872001    .5005351
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est4 stored)

(dropped 10 singleton observations)
(MWFE estimator converged in 27 iterations)

HDFE Linear regression                            Number of obs   =     14,894
Absorbing 4 HDFE groups                           F(   3,     47) =       6.01
Statistics robust to heteroskedasticity           Prob > F        =     0.0015
                                                  R-squared       =     0.0394
                                                  Adj R-squared   =    -0.1593
Number of clusters (state)   =         50         Within R-sq.    =     0.0012
Number of clusters (year)    =         48         Root MSE        =     0.3674

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_cit~ss_m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .0953393   .0279567     3.41   0.001     .0390977    .1515809
 ptonpselect |   .0967933   .0445479     2.17   0.035     .0071744    .1864122
 nptomselect |  -.0895443   .1145636    -0.78   0.438    -.3200165    .1409279
       _cons |   .4934646   .0026816   184.02   0.000       .48807    .4988593
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est5 stored)

(dropped 10 singleton observations)
(MWFE estimator converged in 27 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.

HDFE Linear regression                            Number of obs   =     14,894
Absorbing 4 HDFE groups                           F(   3,     47) =       5.66
Statistics robust to heteroskedasticity           Prob > F        =     0.0021
                                                  R-squared       =     0.0401
                                                  Adj R-squared   =    -0.1583
Number of clusters (state)   =         50         Within R-sq.    =     0.0015
Number of clusters (year)    =         48         Root MSE        =     0.3671

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_ci~ote_m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .1305399   .0324927     4.02   0.000     .0651731    .1959067
 ptonpselect |   .0821508   .0704998     1.17   0.250    -.0596765    .2239781
 nptomselect |  -.0453919   .1045265    -0.43   0.666    -.2556722    .1648884
       _cons |   .4907239   .0030101   163.02   0.000     .4846683    .4967795
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est6 stored)

(dropped 10 singleton observations)
(MWFE estimator converged in 27 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.

HDFE Linear regression                            Number of obs   =     14,894
Absorbing 4 HDFE groups                           F(   3,     47) =       4.02
Statistics robust to heteroskedasticity           Prob > F        =     0.0126
                                                  R-squared       =     0.0272
                                                  Adj R-squared   =    -0.1740
Number of clusters (state)   =         50         Within R-sq.    =     0.0006
Number of clusters (year)    =         48         Root MSE        =     0.3694

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_ci~ate_m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .0841509   .0320227     2.63   0.012     .0197296    .1485722
 ptonpselect |    .053062   .0325342     1.63   0.110    -.0123883    .1185123
 nptomselect |  -.0332587   .0751484    -0.44   0.660    -.1844378    .1179203
       _cons |   .4944763    .002574   192.11   0.000     .4892982    .4996545
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est7 stored)
In [193]:
%html
esttab, r2 se star(+ 0.10 * 0.05 ** 0.01) drop(   _cons) html
esttab using "1-select/2-outcome-specs-stars.csv", replace r2 se star(+ 0.10 * 0.05 ** 0.01) drop(   _cons)

(1) (2) (3) (4) (5) (6) (7)
cites_pos_m cites_pos_m_log pct_cites10_pos_m pct_cites_all_m pct_cites_discuss_m pct_cites_quote_m pct_cites_outstate_m

ptomselect 0.887+ 0.0654+ 0.0767+ 0.0943* 0.0953** 0.131** 0.0842*
(0.499) (0.0371) (0.0444) (0.0356) (0.0280) (0.0325) (0.0320)
 
ptonpselect 0.650 0.0814+ 0.0574 0.0572 0.0968* 0.0822 0.0531
(1.018) (0.0452) (0.0424) (0.0537) (0.0445) (0.0705) (0.0325)
 
nptomselect -0.650 -0.0519 -0.103 -0.0838 -0.0895 -0.0454 -0.0333
(1.910) (0.101) (0.115) (0.141) (0.115) (0.105) (0.0751)

N 14894 14894 14894 14894 14894 14894 14894
R2 0.618 0.807 0.043 0.045 0.039 0.040 0.027

Standard errors in parentheses
+ p < 0.10, * p < 0.05, ** p < 0.01
(output written to 1-select/2-outcome-specs-stars.csv)

Effect on Caseload

In [194]:
eststo clear

* + main reg with state case controls
eststo: reghdfe pct_cites_pos_m  ptomselect ptonpselect nptomselect  , absorb(stateyear  x_decade_start birth_decade statefe##c.(x_decade_start casetype_*_m pca_* pct_numops) year#c.(casetype_*_m pca_* pct_numops) ) cluster(state year)

* P-to-M effect holds under random assignment
eststo: reghdfe pct_cites_pos_m  ptomselect ptonpselect nptomselect if rule_rand_assign, absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) ) cluster(state year)

* P-to-NP increases caseload
eststo: reghdfe pct_numops  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) ) cluster(state year)

* effect on proportion criminal cases
eststo: reghdfe pct_casetype_crim_m  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) ) cluster(state year)

* effect on case-predicted quality
eststo: reghdfe pred_pct_cites_pos_m  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) ) cluster(state year)

(dropped 10 singleton observations)
(MWFE estimator converged in 88 iterations)

HDFE Linear regression                            Number of obs   =     14,894
Absorbing 5 HDFE groups                           F(   3,     47) =       3.01
Statistics robust to heteroskedasticity           Prob > F        =     0.0395
                                                  R-squared       =     0.1634
                                                  Adj R-squared   =    -0.0985
Number of clusters (state)   =         50         Within R-sq.    =     0.0009
Number of clusters (year)    =         48         Root MSE        =     0.3577

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_~s_pos_m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .0902288   .0413153     2.18   0.034     .0071131    .1733445
 ptonpselect |   .0870496    .047807     1.82   0.075    -.0091257    .1832249
 nptomselect |  -.0479332   .1483774    -0.32   0.748      -.34643    .2505636
       _cons |   .4931823   .0034208   144.17   0.000     .4863005     .500064
------------------------------------------------------------------------------

Absorbed degrees of freedom:
--------------------------------------------------------------------+
                Absorbed FE | Categories  - Redundant  = Num. Coefs |
----------------------------+---------------------------------------|
                stateyearfe |      2428        2428           0    *|
             x_decade_start |         9           0           9     |
               birth_decade |        10           1           9     |
                    statefe |        52          52           0    *|
   statefe#c.x_decade_start |        52           0          52    ?|
 statefe#c.casetype_admin_m |        52           0          52    ?|
   statefe#c.casetype_civ_m |        52           1          51    ?|
   statefe#c.casetype_con_m |        52           0          52    ?|
  statefe#c.casetype_crim_m |        52           0          52    ?|
            statefe#c.pca_1 |        52           0          52    ?|
            statefe#c.pca_2 |        52           0          52    ?|
            statefe#c.pca_3 |        52           0          52    ?|
            statefe#c.pca_4 |        52           0          52    ?|
            statefe#c.pca_5 |        52           0          52    ?|
       statefe#c.pct_numops |        52           0          52    ?|
    year#c.casetype_admin_m |        48           0          48    ?|
      year#c.casetype_civ_m |        48           0          48    ?|
      year#c.casetype_con_m |        48           0          48    ?|
     year#c.casetype_crim_m |        48           0          48    ?|
               year#c.pca_1 |        48           0          48    ?|
               year#c.pca_2 |        48           0          48    ?|
               year#c.pca_3 |        48           0          48    ?|
               year#c.pca_4 |        48           0          48    ?|
               year#c.pca_5 |        48           0          48    ?|
          year#c.pct_numops |        48           0          48    ?|
--------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est1 stored)

(MWFE estimator converged in 24 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.
warning: missing F statistic; dropped variables due to collinearity or too few c
> lusters

HDFE Linear regression                            Number of obs   =     10,833
Absorbing 4 HDFE groups                           F(   3,     34) =          .
Statistics robust to heteroskedasticity           Prob > F        =          .
                                                  R-squared       =     0.0460
                                                  Adj R-squared   =    -0.1491
Number of clusters (state)   =         35         Within R-sq.    =     0.0012
Number of clusters (year)    =         48         Root MSE        =     0.3650

                            (Std. Err. adjusted for 35 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_~s_pos_m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .1222301   .0306067     3.99   0.000     .0600297    .1844305
 ptonpselect |   .0513521   .0538121     0.95   0.347    -.0580073    .1607114
 nptomselect |  -.0563217   .0506607    -1.11   0.274    -.1592765    .0466332
       _cons |   .4926745    .002021   243.77   0.000     .4885672    .4967817
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      1745        1745           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        37          37           0    *|
 statefe#c.x_decade_start |        37           0          37    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est2 stored)

(dropped 10 singleton observations)
(MWFE estimator converged in 27 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.

HDFE Linear regression                            Number of obs   =     14,894
Absorbing 4 HDFE groups                           F(   3,     47) =       4.97
Statistics robust to heteroskedasticity           Prob > F        =     0.0044
                                                  R-squared       =     0.0275
                                                  Adj R-squared   =    -0.1736
Number of clusters (state)   =         50         Within R-sq.    =     0.0001
Number of clusters (year)    =         48         Root MSE        =     0.3720

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
  pct_numops |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .0050735   .0401327     0.13   0.900    -.0756631    .0858101
 ptonpselect |   .0681291   .0195106     3.49   0.001     .0288788    .1073794
 nptomselect |   .0465129   .1279355     0.36   0.718    -.2108601    .3038859
       _cons |   .5075299   .0029254   173.49   0.000     .5016448     .513415
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est3 stored)

(dropped 10 singleton observations)
(MWFE estimator converged in 27 iterations)

HDFE Linear regression                            Number of obs   =     14,894
Absorbing 4 HDFE groups                           F(   3,     47) =       0.99
Statistics robust to heteroskedasticity           Prob > F        =     0.4071
                                                  R-squared       =     0.0606
                                                  Adj R-squared   =    -0.1337
Number of clusters (state)   =         50         Within R-sq.    =     0.0002
Number of clusters (year)    =         48         Root MSE        =     0.3644

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_casety~m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |  -.0303863    .023947    -1.27   0.211    -.0785614    .0177888
 ptonpselect |  -.0508411    .071966    -0.71   0.483     -.195618    .0939358
 nptomselect |  -.0344955    .061858    -0.56   0.580    -.1589378    .0899467
       _cons |    .517894   .0020262   255.59   0.000     .5138177    .5219702
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est4 stored)

(dropped 16 singleton observations)
(MWFE estimator converged in 27 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.
warning: missing F statistic; dropped variables due to collinearity or too few c
> lusters

HDFE Linear regression                            Number of obs   =     14,894
Absorbing 4 HDFE groups                           F(   3,     47) =          .
Statistics robust to heteroskedasticity           Prob > F        =          .
                                                  R-squared       =     0.6649
                                                  Adj R-squared   =     0.5956
Number of clusters (state)   =         50         Within R-sq.    =     0.0004
Number of clusters (year)    =         48         Root MSE        =     0.0538

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pred_pct_c~m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |  -.0029266   .0040603    -0.72   0.475    -.0110949    .0052418
 ptonpselect |   .0035168   .0012628     2.79   0.008     .0009765    .0060572
 nptomselect |  -.0146296   .0158121    -0.93   0.360    -.0464394    .0171802
       _cons |   .5009315   .0003758  1332.88   0.000     .5001754    .5016876
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est5 stored)
In [195]:
%html
esttab, r2 se star(+ 0.10 * 0.05 ** 0.01) drop(   _cons) html
esttab using "1-select/3-caseload-stars.csv", replace r2 se star(+ 0.10 * 0.05 ** 0.01) drop(   _cons)

(1) (2) (3) (4) (5)
pct_cites_pos_m pct_cites_pos_m pct_numops pct_casetype_crim_m pred_pct_cites_pos_m

ptomselect 0.0902* 0.122** 0.00507 -0.0304 -0.00293
(0.0413) (0.0306) (0.0401) (0.0239) (0.00406)
 
ptonpselect 0.0870+ 0.0514 0.0681** -0.0508 0.00352**
(0.0478) (0.0538) (0.0195) (0.0720) (0.00126)
 
nptomselect -0.0479 -0.0563 0.0465 -0.0345 -0.0146
(0.148) (0.0507) (0.128) (0.0619) (0.0158)

N 14894 10833 14894 14894 14894
R2 0.163 0.046 0.028 0.061 0.665

Standard errors in parentheses
+ p < 0.10, * p < 0.05, ** p < 0.01
(output written to 1-select/3-caseload-stars.csv)

Effect on Judge Types

In [196]:
eststo clear

eststo: reghdfe ptomselect x_partisan x_startage x_female x_topschool x_prevjudge x_prevprivate x_prevprof [aw=inv_length], absorb(stateyear one#c.(x_startage_missing x_female_missing x_topschool_missing x_prevjudge_missing x_prevprof_missing x_prevpolitician_missing x_prevprivate_missing))  cluster(stateyear) nocons
eststo: reghdfe ptonpselect x_partisan x_startage x_female x_topschool x_prevjudge x_prevprivate x_prevprof [aw=inv_length], absorb(stateyear one#c.(x_startage_missing x_female_missing x_topschool_missing x_prevjudge_missing x_prevprof_missing x_prevpolitician_missing x_prevprivate_missing))  cluster(stateyear) nocons
eststo: reghdfe nptomselect x_partisan x_startage x_female x_topschool x_prevjudge x_prevprivate x_prevprof [aw=inv_length], absorb(stateyear one#c.(x_startage_missing x_female_missing x_topschool_missing x_prevjudge_missing x_prevprof_missing x_prevpolitician_missing x_prevprivate_missing))  cluster(stateyear) nocons

eststo: reghdfe pct_cites_outstate_m x_partisan x_startage x_female x_topschool x_prevjudge x_prevprivate x_prevprof pred_pct_cites_pos_m if non_reform , absorb(stateyear  x_decade_start birth_decade statefe##c.(x_decade_start)  one#c.(x_startage_missing x_female_missing x_topschool_missing x_prevjudge_missing x_prevprof_missing x_prevpolitician_missing x_prevprivate_missing))  cluster(stateyear) nocons

(dropped 8 singleton observations)
(MWFE estimator converged in 8 iterations)

HDFE Linear regression                            Number of obs   =     14,983
Absorbing 2 HDFE groups                           F(   7,   2437) =      12.65
Statistics robust to heteroskedasticity           Prob > F        =     0.0000
                                                  R-squared       =     0.8840
                                                  Adj R-squared   =     0.8614
                                                  Within R-sq.    =     0.0128
Number of clusters (stateyearfe) =      2,438     Root MSE        =     0.0963

                         (Std. Err. adjusted for 2,438 clusters in stateyearfe)
-------------------------------------------------------------------------------
              |               Robust
   ptomselect |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
--------------+----------------------------------------------------------------
   x_partisan |  -.0198676   .0023985    -8.28   0.000    -.0245709   -.0151643
   x_startage |   .0010937   .0001358     8.05   0.000     .0008274      .00136
     x_female |   .0093351   .0021172     4.41   0.000     .0051834    .0134868
  x_topschool |   .0037141   .0015834     2.35   0.019     .0006092     .006819
  x_prevjudge |   6.10e-07   .0022905     0.00   1.000    -.0044909    .0044921
x_prevprivate |   .0000986   .0019452     0.05   0.960    -.0037157     .003913
   x_prevprof |   .0006603   .0016017     0.41   0.680    -.0024804     .003801
-------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------------+
                    Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------------+---------------------------------------|
                    stateyearfe |      2438        2438           0    *|
       one#c.x_startage_missing |         1           0           1    ?|
         one#c.x_female_missing |         1           1           0    ?|
      one#c.x_topschool_missing |         1           0           1    ?|
      one#c.x_prevjudge_missing |         1           0           1    ?|
       one#c.x_prevprof_missing |         1           0           1    ?|
 one#c.x_prevpolitician_missing |         1           0           1    ?|
    one#c.x_prevprivate_missing |         1           0           1    ?|
------------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est1 stored)

(dropped 8 singleton observations)
(MWFE estimator converged in 8 iterations)

HDFE Linear regression                            Number of obs   =     14,983
Absorbing 2 HDFE groups                           F(   7,   2437) =       2.51
Statistics robust to heteroskedasticity           Prob > F        =     0.0141
                                                  R-squared       =     0.9165
                                                  Adj R-squared   =     0.9002
                                                  Within R-sq.    =     0.0031
Number of clusters (stateyearfe) =      2,438     Root MSE        =     0.0492

                         (Std. Err. adjusted for 2,438 clusters in stateyearfe)
-------------------------------------------------------------------------------
              |               Robust
  ptonpselect |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
--------------+----------------------------------------------------------------
   x_partisan |  -.0052207   .0015278    -3.42   0.001    -.0082166   -.0022249
   x_startage |   -.000254   .0000759    -3.35   0.001    -.0004029   -.0001051
     x_female |  -.0015802   .0004952    -3.19   0.001    -.0025512   -.0006091
  x_topschool |   .0004184   .0004813     0.87   0.385    -.0005253    .0013622
  x_prevjudge |  -.0004427   .0002408    -1.84   0.066    -.0009149    .0000295
x_prevprivate |  -.0015001   .0007141    -2.10   0.036    -.0029003   -.0000998
   x_prevprof |  -.0010575   .0010517    -1.01   0.315    -.0031197    .0010047
-------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------------+
                    Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------------+---------------------------------------|
                    stateyearfe |      2438        2438           0    *|
       one#c.x_startage_missing |         1           0           1    ?|
         one#c.x_female_missing |         1           1           0    ?|
      one#c.x_topschool_missing |         1           0           1    ?|
      one#c.x_prevjudge_missing |         1           0           1    ?|
       one#c.x_prevprof_missing |         1           0           1    ?|
 one#c.x_prevpolitician_missing |         1           0           1    ?|
    one#c.x_prevprivate_missing |         1           0           1    ?|
------------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est2 stored)

(dropped 8 singleton observations)
(MWFE estimator converged in 8 iterations)

HDFE Linear regression                            Number of obs   =     14,983
Absorbing 2 HDFE groups                           F(   7,   2437) =       6.60
Statistics robust to heteroskedasticity           Prob > F        =     0.0000
                                                  R-squared       =     0.8411
                                                  Adj R-squared   =     0.8100
                                                  Within R-sq.    =     0.0073
Number of clusters (stateyearfe) =      2,438     Root MSE        =     0.0678

                         (Std. Err. adjusted for 2,438 clusters in stateyearfe)
-------------------------------------------------------------------------------
              |               Robust
  nptomselect |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
--------------+----------------------------------------------------------------
   x_partisan |   .0018537   .0019377     0.96   0.339    -.0019459    .0056534
   x_startage |   .0005835   .0001025     5.69   0.000     .0003825    .0007845
     x_female |   .0132441   .0047375     2.80   0.005     .0039541    .0225341
  x_topschool |   .0025869   .0014264     1.81   0.070    -.0002102     .005384
  x_prevjudge |  -.0003207   .0018001    -0.18   0.859    -.0038506    .0032092
x_prevprivate |   .0097469   .0019259     5.06   0.000     .0059704    .0135235
   x_prevprof |   .0022986     .00192     1.20   0.231    -.0014665    .0060636
-------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------------+
                    Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------------+---------------------------------------|
                    stateyearfe |      2438        2438           0    *|
       one#c.x_startage_missing |         1           0           1    ?|
         one#c.x_female_missing |         1           1           0    ?|
      one#c.x_topschool_missing |         1           0           1    ?|
      one#c.x_prevjudge_missing |         1           0           1    ?|
       one#c.x_prevprof_missing |         1           0           1    ?|
 one#c.x_prevpolitician_missing |         1           0           1    ?|
    one#c.x_prevprivate_missing |         1           0           1    ?|
------------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est3 stored)

(dropped 80 singleton observations)
(MWFE estimator converged in 30 iterations)

HDFE Linear regression                            Number of obs   =     13,371
Absorbing 5 HDFE groups                           F(   8,   2277) =      11.58
Statistics robust to heteroskedasticity           Prob > F        =     0.0000
                                                  R-squared       =     0.0449
                                                  Adj R-squared   =    -0.1653
                                                  Within R-sq.    =     0.0082
Number of clusters (stateyearfe) =      2,278     Root MSE        =     0.3681

                         (Std. Err. adjusted for 2,278 clusters in stateyearfe)
-------------------------------------------------------------------------------
              |               Robust
pct_cit~ate_m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
--------------+----------------------------------------------------------------
   x_partisan |   .0064571   .0086418     0.75   0.455    -.0104895    .0234036
   x_startage |   -.001696   .0009982    -1.70   0.089    -.0036535    .0002614
     x_female |   .0829133   .0231972     3.57   0.000     .0374234    .1284033
  x_topschool |   .0208263   .0112603     1.85   0.065    -.0012553    .0429079
  x_prevjudge |   .0160731   .0093816     1.71   0.087    -.0023242    .0344704
x_prevprivate |   .0313783   .0102133     3.07   0.002     .0113499    .0514067
   x_prevprof |   .0164923   .0144155     1.14   0.253    -.0117767    .0447613
pred_pct_ci~m |   .5049997    .065114     7.76   0.000     .3773106    .6326887
-------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------------+
                    Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------------+---------------------------------------|
                    stateyearfe |      2278        2278           0    *|
                 x_decade_start |         9           0           9     |
                   birth_decade |        10           1           9     |
                        statefe |        52           1          51    ?|
       statefe#c.x_decade_start |        52           0          52    ?|
       one#c.x_startage_missing |         1           0           1    ?|
         one#c.x_female_missing |         1           1           0    ?|
      one#c.x_topschool_missing |         1           0           1    ?|
      one#c.x_prevjudge_missing |         1           0           1    ?|
       one#c.x_prevprof_missing |         1           0           1    ?|
 one#c.x_prevpolitician_missing |         1           0           1    ?|
    one#c.x_prevprivate_missing |         1           0           1    ?|
------------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est4 stored)
In [197]:
%html
esttab, r2 se star(+ 0.10 * 0.05 ** 0.01)  html
esttab using "1-select/4-judge-types-stars.csv", replace r2 se star(+ 0.10 * 0.05 ** 0.01)

(1) (2) (3) (4)
ptomselect ptonpselect nptomselect pct_cites_outstate_m

x_partisan -0.0199** -0.00522** 0.00185 0.00646
(0.00240) (0.00153) (0.00194) (0.00864)
 
x_startage 0.00109** -0.000254** 0.000583** -0.00170+
(0.000136) (0.0000759) (0.000103) (0.000998)
 
x_female 0.00934** -0.00158** 0.0132** 0.0829**
(0.00212) (0.000495) (0.00474) (0.0232)
 
x_topschool 0.00371* 0.000418 0.00259+ 0.0208+
(0.00158) (0.000481) (0.00143) (0.0113)
 
x_prevjudge 0.000000610 -0.000443+ -0.000321 0.0161+
(0.00229) (0.000241) (0.00180) (0.00938)
 
x_prevprivate 0.0000986 -0.00150* 0.00975** 0.0314**
(0.00195) (0.000714) (0.00193) (0.0102)
 
x_prevprof 0.000660 -0.00106 0.00230 0.0165
(0.00160) (0.00105) (0.00192) (0.0144)
 
pred_pct_cites_pos_m 0.505**
(0.0651)

N 14983 14983 14983 13371
R2 0.884 0.917 0.841 0.045

Standard errors in parentheses
+ p < 0.10, * p < 0.05, ** p < 0.01
(output written to 1-select/4-judge-types-stars.csv)

Appendix Results

Additional Summary Statistics

In [198]:
binscatter cites_pos_m_st txt_words_st, title(Standardized by Court-Year) xtitle("Work Output") ytitle("Work Quality") 
graph export 0-summ-stats/A-quality-vs-output-st.pdf, replace
graph display

(file 0-summ-stats/A-quality-vs-output-st.pdf written in PDF format)

In [199]:
binscatter pct_cites_pos x_rating_good if year >= 1988, absorb(stateyear) xtitle("Normalized ABA Rating") ytitle("Work Impact (Total Citations)") 
graph display

Conditional logit regressions of work quality and work output on bar association ratings.

In [200]:
eststo clear
eststo: clogit x_rating_good i.year  pct_cites_pos_m if year >= 1988, group(state) cl(state)
eststo: clogit x_rating_good i.year  pct_txt_words if year >= 1988, group(state) cl(state)
eststo: clogit x_rating_good i.year  pct_cites_pos_m pct_txt_words if year >= 1988, group(state) cl(state)

note: multiple positive outcomes within groups encountered.
note: 6 groups (109 obs) dropped because of all positive or
      all negative outcomes.

Iteration 0:   log pseudolikelihood = -22.577952  
Iteration 1:   log pseudolikelihood = -22.516961  
Iteration 2:   log pseudolikelihood = -22.516752  
Iteration 3:   log pseudolikelihood = -22.516752  

Conditional (fixed-effects) logistic regression

                                                Number of obs     =         61
                                                Wald chi2(1)      =          .
                                                Prob > chi2       =          .
Log pseudolikelihood = -22.516752               Pseudo R2         =     0.2166

                                   (Std. Err. adjusted for 2 clusters in state)
-------------------------------------------------------------------------------
              |               Robust
x_rating_good |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
--------------+----------------------------------------------------------------
         year |
        1989  |  -1.007578   1.223537    -0.82   0.410    -3.405667     1.39051
        1990  |  -.7706784   2.233284    -0.35   0.730    -5.147834    3.606477
        1991  |   .0234976   2.466314     0.01   0.992    -4.810389    4.857385
        1992  |  -.5761852   2.957905    -0.19   0.846    -6.373572    5.221201
        1993  |   -.070244   3.522161    -0.02   0.984    -6.973553    6.833066
        1994  |  -.5168791   2.669741    -0.19   0.846    -5.749476    4.715718
              |
pct_cites_p~m |   3.553467   1.208791     2.94   0.003      1.18428    5.922654
-------------------------------------------------------------------------------
(est1 stored)

note: multiple positive outcomes within groups encountered.
note: 6 groups (109 obs) dropped because of all positive or
      all negative outcomes.

Iteration 0:   log pseudolikelihood = -27.033547  
Iteration 1:   log pseudolikelihood = -26.968883  
Iteration 2:   log pseudolikelihood = -26.968806  
Iteration 3:   log pseudolikelihood = -26.968806  

Conditional (fixed-effects) logistic regression

                                                Number of obs     =         61
                                                Wald chi2(1)      =          .
                                                Prob > chi2       =          .
Log pseudolikelihood = -26.968806               Pseudo R2         =     0.0617

                                   (Std. Err. adjusted for 2 clusters in state)
-------------------------------------------------------------------------------
              |               Robust
x_rating_good |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
--------------+----------------------------------------------------------------
         year |
        1989  |  -.3573492   1.260036    -0.28   0.777    -2.826974    2.112276
        1990  |  -.1796593   1.284479    -0.14   0.889    -2.697191    2.337873
        1991  |  -.0373274   1.504137    -0.02   0.980    -2.985381    2.910726
        1992  |  -.5041772   1.747372    -0.29   0.773    -3.928963    2.920609
        1993  |  -.0359095    2.62418    -0.01   0.989    -5.179208    5.107389
        1994  |  -.4857259   1.561516    -0.31   0.756    -3.546242     2.57479
              |
pct_txt_words |   1.721709   1.015963     1.69   0.090    -.2695406    3.712959
-------------------------------------------------------------------------------
(est2 stored)

note: multiple positive outcomes within groups encountered.
note: 6 groups (109 obs) dropped because of all positive or
      all negative outcomes.

Iteration 0:   log pseudolikelihood = -22.489342  
Iteration 1:   log pseudolikelihood = -22.420091  
Iteration 2:   log pseudolikelihood = -22.419897  
Iteration 3:   log pseudolikelihood = -22.419897  

Conditional (fixed-effects) logistic regression

                                                Number of obs     =         61
                                                Wald chi2(1)      =          .
                                                Prob > chi2       =          .
Log pseudolikelihood = -22.419897               Pseudo R2         =     0.2200

                                   (Std. Err. adjusted for 2 clusters in state)
-------------------------------------------------------------------------------
              |               Robust
x_rating_good |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
--------------+----------------------------------------------------------------
         year |
        1989  |  -.9578232   1.352947    -0.71   0.479    -3.609551    1.693904
        1990  |  -.7282315   2.219071    -0.33   0.743     -5.07753    3.621067
        1991  |   .0695817   2.485477     0.03   0.978    -4.801864    4.941028
        1992  |  -.5807336   2.925253    -0.20   0.843    -6.314125    5.152657
        1993  |  -.0628804   3.546632    -0.02   0.986    -7.014152    6.888391
        1994  |  -.5075479   2.675466    -0.19   0.850    -5.751365    4.736269
              |
pct_cites_p~m |   3.331457   .9013418     3.70   0.000      1.56486    5.098055
pct_txt_words |   .5006724   .5530986     0.91   0.365    -.5833809    1.584726
-------------------------------------------------------------------------------
(est3 stored)
In [201]:
%html
esttab, ar2 se star(+ 0.10 * 0.05 ** 0.01) drop(*year*) html
esttab using 0-summ-stats/bar-evals.csv, ar2 se star(+ 0.10 * 0.05 ** 0.01) drop(*year*) replace

(1) (2) (3)
x_rating_good x_rating_good x_rating_good

x_rating_good
pct_cites_pos_m 3.553** 3.331**
(1.209) (0.901)
 
pct_txt_words 1.722+ 0.501
(1.016) (0.553)

N 61 61 61
adj. R2

Standard errors in parentheses
+ p < 0.10, * p < 0.05, ** p < 0.01
(output written to 0-summ-stats/bar-evals.csv)

Coefficient Plot for Main Results

In [202]:
eststo clear
eststo: reghdfe pct_cites_pos_m  ptomselect ptonpselect nptomselect  , absorb(stateyear  ) cluster(state year)  
eststo: reghdfe pct_txt_words  ptomselect ptonpselect nptomselect  , absorb(stateyear  ) cluster(state year)  nocons

(MWFE estimator converged in 1 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.

HDFE Linear regression                            Number of obs   =     14,996
Absorbing 1 HDFE group                            F(   3,     47) =       7.80
Statistics robust to heteroskedasticity           Prob > F        =     0.0003
                                                  R-squared       =     0.0040
                                                  Adj R-squared   =    -0.1896
Number of clusters (state)   =         50         Within R-sq.    =     0.0039
Number of clusters (year)    =         48         Root MSE        =     0.3722

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_~s_pos_m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .1776986   .0429922     4.13   0.000     .0912094    .2641878
 ptonpselect |   .1010445   .0476802     2.12   0.039     .0051242    .1969647
 nptomselect |  -.0686153   .1364205    -0.50   0.617    -.3430579    .2058273
       _cons |   .4878864   .0033343   146.32   0.000     .4811787    .4945941
------------------------------------------------------------------------------

Absorbed degrees of freedom:
-----------------------------------------------------+
 Absorbed FE | Categories  - Redundant  = Num. Coefs |
-------------+---------------------------------------|
 stateyearfe |      2438        2438           0    *|
-----------------------------------------------------+
* = FE nested within cluster; treated as redundant for DoF computation
(est1 stored)

(MWFE estimator converged in 1 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.
warning: missing F statistic; dropped variables due to collinearity or too few c
> lusters

HDFE Linear regression                            Number of obs   =     14,996
Absorbing 1 HDFE group                            F(   3,     47) =          .
Statistics robust to heteroskedasticity           Prob > F        =          .
                                                  R-squared       =     0.0008
                                                  Adj R-squared   =    -0.1934
Number of clusters (state)   =         50         Within R-sq.    =     0.0008
Number of clusters (year)    =         48         Root MSE        =     0.3730

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_txt_wo~s |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .0631543   .0348467     1.81   0.076    -.0069483    .1332569
 ptonpselect |   .0886494    .013895     6.38   0.000     .0606963    .1166026
 nptomselect |  -.0628946   .0732454    -0.86   0.395    -.2102453    .0844561
------------------------------------------------------------------------------

Absorbed degrees of freedom:
-----------------------------------------------------+
 Absorbed FE | Categories  - Redundant  = Num. Coefs |
-------------+---------------------------------------|
 stateyearfe |      2438        2438           0    *|
-----------------------------------------------------+
* = FE nested within cluster; treated as redundant for DoF computation
(est2 stored)
In [203]:
coefplot est1 est2, drop(_cons) xline(0, lpattern(dash) lcolor(gray)) rename(ptonpselect = "Partisan to Non-Partisan" ptomselect = "Partisan to Merit" nptomselect = "Non-Partisan to Merit") legend(order(2 "Work Quality" 4 "Work Output" )) xsize(8)
graph export 1-select/main-coefplot-1.pdf, replace


(file 1-select/main-coefplot-1.pdf written in PDF format)
In [204]:
eststo clear
eststo: reghdfe pct_cites_pos_m  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) xt_exper) cluster(state year)
eststo: reghdfe pct_txt_words  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) xt_exper) cluster(state year) nocons

(dropped 14 singleton observations)
(MWFE estimator converged in 31 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.

HDFE Linear regression                            Number of obs   =     14,890
Absorbing 5 HDFE groups                           F(   3,     47) =       2.59
Statistics robust to heteroskedasticity           Prob > F        =     0.0639
                                                  R-squared       =     0.0512
                                                  Adj R-squared   =    -0.1481
Number of clusters (state)   =         50         Within R-sq.    =     0.0011
Number of clusters (year)    =         48         Root MSE        =     0.3656

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_~s_pos_m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |    .083149   .0408577     2.04   0.048     .0009539    .1653441
 ptonpselect |   .0866382   .0472728     1.83   0.073    -.0084624    .1817389
 nptomselect |   -.106856   .1553657    -0.69   0.495    -.4194115    .2056995
       _cons |   .4945942   .0036461   135.65   0.000     .4872592    .5019292
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
                 xt_exper |        34           1          33    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est1 stored)

(dropped 14 singleton observations)
(MWFE estimator converged in 31 iterations)

HDFE Linear regression                            Number of obs   =     14,890
Absorbing 5 HDFE groups                           F(   3,     47) =       2.51
Statistics robust to heteroskedasticity           Prob > F        =     0.0704
                                                  R-squared       =     0.0585
                                                  Adj R-squared   =    -0.1393
Number of clusters (state)   =         50         Within R-sq.    =     0.0004
Number of clusters (year)    =         48         Root MSE        =     0.3642

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_txt_wo~s |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .0279854   .0421183     0.66   0.510    -.0567458    .1127165
 ptonpselect |   .1000869   .0476567     2.10   0.041      .004214    .1959599
 nptomselect |  -.0717419   .1308154    -0.55   0.586    -.3349086    .1914248
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
                 xt_exper |        34           1          33    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est2 stored)
In [205]:
coefplot est1 est2, drop(_cons) xline(0, lpattern(dash) lcolor(gray)) rename(ptonpselect = "Partisan to Non-Partisan" ptomselect = "Partisan to Merit" nptomselect = "Non-Partisan to Merit") legend(order(2 "Work Qualtiy" 4 "Work Output" ))
graph export 1-select/main-coefplot-2.pdf, replace


(file 1-select/main-coefplot-2.pdf written in PDF format)
In [206]:
reghdfe pct_cites_pos_m  npselect mselect , absorb(stateyear) cluster(state year)
(MWFE estimator converged in 1 iterations)

HDFE Linear regression                            Number of obs   =     14,996
Absorbing 1 HDFE group                            F(   2,     47) =       4.50
Statistics robust to heteroskedasticity           Prob > F        =     0.0162
                                                  R-squared       =     0.0037
                                                  Adj R-squared   =    -0.1899
Number of clusters (state)   =         50         Within R-sq.    =     0.0036
Number of clusters (year)    =         48         Root MSE        =     0.3723

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_~s_pos_m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
    npselect |   .1101479   .0508811     2.16   0.036     .0077884    .2125074
     mselect |    .119597   .0533553     2.24   0.030       .01226     .226934
       _cons |   .4654768   .0119954    38.80   0.000     .4413452    .4896084
------------------------------------------------------------------------------

Absorbed degrees of freedom:
-----------------------------------------------------+
 Absorbed FE | Categories  - Redundant  = Num. Coefs |
-------------+---------------------------------------|
 stateyearfe |      2438        2438           0    *|
-----------------------------------------------------+
* = FE nested within cluster; treated as redundant for DoF computation

Additional Quality Variables (Experience FE)

In [207]:
* cohort FE / trends spec
eststo clear

eststo: reghdfe cites_pos_m  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) xt_exper ) cluster(state year)
eststo: reghdfe cites_pos_m_log  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) xt_exper ) cluster(state year)
eststo: reghdfe pct_cites10_pos_m  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) xt_exper ) cluster(state year)

eststo: reghdfe pct_cites_all_m  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) xt_exper ) cluster(state year)
eststo: reghdfe pct_cites_discuss_m  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) xt_exper ) cluster(state year)
eststo: reghdfe pct_cites_quote_m  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) xt_exper ) cluster(state year)
eststo: reghdfe pct_cites_outstate_m  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) xt_exper ) cluster(state year)

(dropped 20 singleton observations)
(MWFE estimator converged in 31 iterations)

HDFE Linear regression                            Number of obs   =     14,890
Absorbing 5 HDFE groups                           F(   3,     47) =       0.95
Statistics robust to heteroskedasticity           Prob > F        =     0.4222
                                                  R-squared       =     0.6187
                                                  Adj R-squared   =     0.5386
Number of clusters (state)   =         50         Within R-sq.    =     0.0001
Number of clusters (year)    =         48         Root MSE        =     8.5236

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
 cites_pos_m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .7511501   .5097184     1.47   0.147     -.274271    1.776571
 ptonpselect |   .4639358   1.081641     0.43   0.670    -1.712045    2.639917
 nptomselect |  -.8730786   1.916998    -0.46   0.651    -4.729581    2.983423
       _cons |   13.07132   .0497013   263.00   0.000     12.97133    13.17131
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
                 xt_exper |        34           1          33    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est1 stored)

(dropped 20 singleton observations)
(MWFE estimator converged in 31 iterations)

HDFE Linear regression                            Number of obs   =     14,890
Absorbing 5 HDFE groups                           F(   3,     47) =       1.55
Statistics robust to heteroskedasticity           Prob > F        =     0.2135
                                                  R-squared       =     0.8076
                                                  Adj R-squared   =     0.7672
Number of clusters (state)   =         50         Within R-sq.    =     0.0006
Number of clusters (year)    =         48         Root MSE        =     0.3204

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
~s_pos_m_log |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .0532333   .0371111     1.43   0.158    -.0214246    .1278912
 ptonpselect |   .0758439   .0471017     1.61   0.114    -.0189125    .1706003
 nptomselect |  -.0656928   .1026325    -0.64   0.525    -.2721626    .1407771
       _cons |   2.421446   .0030637   790.36   0.000     2.415282    2.427609
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
                 xt_exper |        34           1          33    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est2 stored)

(dropped 14 singleton observations)
(MWFE estimator converged in 31 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.

HDFE Linear regression                            Number of obs   =     14,890
Absorbing 5 HDFE groups                           F(   3,     47) =       1.21
Statistics robust to heteroskedasticity           Prob > F        =     0.3172
                                                  R-squared       =     0.0490
                                                  Adj R-squared   =    -0.1509
Number of clusters (state)   =         50         Within R-sq.    =     0.0008
Number of clusters (year)    =         48         Root MSE        =     0.3660

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_~0_pos_m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .0568424   .0451131     1.26   0.214    -.0339134    .1475983
 ptonpselect |   .0508266   .0438854     1.16   0.253    -.0374594    .1391127
 nptomselect |   -.124764   .1161475    -1.07   0.288    -.3584226    .1088947
       _cons |   .4974468   .0034376   144.71   0.000     .4905311    .5043624
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
                 xt_exper |        34           1          33    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est3 stored)

(dropped 14 singleton observations)
(MWFE estimator converged in 31 iterations)

HDFE Linear regression                            Number of obs   =     14,890
Absorbing 5 HDFE groups                           F(   3,     47) =       2.30
Statistics robust to heteroskedasticity           Prob > F        =     0.0899
                                                  R-squared       =     0.0512
                                                  Adj R-squared   =    -0.1481
Number of clusters (state)   =         50         Within R-sq.    =     0.0009
Number of clusters (year)    =         48         Root MSE        =     0.3656

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_cite~l_m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .0802583   .0350023     2.29   0.026     .0098427    .1506739
 ptonpselect |   .0467486   .0548216     0.85   0.398    -.0635383    .1570354
 nptomselect |  -.1008366     .14317    -0.70   0.485    -.3888575    .1871843
       _cons |   .4953177   .0033798   146.55   0.000     .4885185    .5021169
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
                 xt_exper |        34           1          33    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est4 stored)

(dropped 14 singleton observations)
(MWFE estimator converged in 31 iterations)

HDFE Linear regression                            Number of obs   =     14,890
Absorbing 5 HDFE groups                           F(   3,     47) =       5.74
Statistics robust to heteroskedasticity           Prob > F        =     0.0020
                                                  R-squared       =     0.0428
                                                  Adj R-squared   =    -0.1584
Number of clusters (state)   =         50         Within R-sq.    =     0.0011
Number of clusters (year)    =         48         Root MSE        =     0.3672

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_cit~ss_m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .0863214    .026469     3.26   0.002     .0330726    .1395701
 ptonpselect |   .0972591   .0376575     2.58   0.013     .0215021    .1730161
 nptomselect |  -.0955546   .1156945    -0.83   0.413     -.328302    .1371928
       _cons |   .4941516   .0026477   186.64   0.000     .4888251     .499478
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
                 xt_exper |        34           1          33    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est5 stored)

(dropped 14 singleton observations)
(MWFE estimator converged in 31 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.

HDFE Linear regression                            Number of obs   =     14,890
Absorbing 5 HDFE groups                           F(   3,     47) =       4.80
Statistics robust to heteroskedasticity           Prob > F        =     0.0053
                                                  R-squared       =     0.0447
                                                  Adj R-squared   =    -0.1561
Number of clusters (state)   =         50         Within R-sq.    =     0.0013
Number of clusters (year)    =         48         Root MSE        =     0.3668

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_ci~ote_m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .1162699    .032115     3.62   0.001     .0516628    .1808769
 ptonpselect |   .0817704   .0669902     1.22   0.228    -.0529964    .2165372
 nptomselect |  -.0580544   .1053691    -0.55   0.584    -.2700296    .1539208
       _cons |   .4918443   .0029734   165.42   0.000     .4858627     .497826
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
                 xt_exper |        34           1          33    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est6 stored)

(dropped 14 singleton observations)
(MWFE estimator converged in 31 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.

HDFE Linear regression                            Number of obs   =     14,890
Absorbing 5 HDFE groups                           F(   3,     47) =       4.20
Statistics robust to heteroskedasticity           Prob > F        =     0.0104
                                                  R-squared       =     0.0319
                                                  Adj R-squared   =    -0.1714
Number of clusters (state)   =         50         Within R-sq.    =     0.0006
Number of clusters (year)    =         48         Root MSE        =     0.3690

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_ci~ate_m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .0758188   .0304107     2.49   0.016     .0146404    .1369973
 ptonpselect |    .063658   .0289333     2.20   0.033     .0054517    .1218643
 nptomselect |  -.0345072   .0754708    -0.46   0.650    -.1863349    .1173204
       _cons |   .4947967   .0024895   198.75   0.000     .4897884    .4998049
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
                 xt_exper |        34           1          33    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est7 stored)
In [208]:
%html
esttab, r2 se star(+ 0.10 * 0.05 ** 0.01) drop(   _cons) html
esttab using "1-select/A-quality-specs-exper.csv", replace r2 se star(+ 0.10 * 0.05 ** 0.01) drop(   _cons)

(1) (2) (3) (4) (5) (6) (7)
cites_pos_m cites_pos_m_log pct_cites10_pos_m pct_cites_all_m pct_cites_discuss_m pct_cites_quote_m pct_cites_outstate_m

ptomselect 0.751 0.0532 0.0568 0.0803* 0.0863** 0.116** 0.0758*
(0.510) (0.0371) (0.0451) (0.0350) (0.0265) (0.0321) (0.0304)
 
ptonpselect 0.464 0.0758 0.0508 0.0467 0.0973* 0.0818 0.0637*
(1.082) (0.0471) (0.0439) (0.0548) (0.0377) (0.0670) (0.0289)
 
nptomselect -0.873 -0.0657 -0.125 -0.101 -0.0956 -0.0581 -0.0345
(1.917) (0.103) (0.116) (0.143) (0.116) (0.105) (0.0755)

N 14890 14890 14890 14890 14890 14890 14890
R2 0.619 0.808 0.049 0.051 0.043 0.045 0.032

Standard errors in parentheses
+ p < 0.10, * p < 0.05, ** p < 0.01
(output written to 1-select/A-quality-specs-exper.csv)

Additional Measures of Output

In [209]:
eststo clear

eststo: reghdfe txt_words  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) ) cluster(state year)
eststo: reghdfe txt_words_log  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) ) cluster(state year)
eststo: reghdfe pct_txt_chars  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) ) cluster(state year)
eststo: reghdfe pct_txt_sents  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) ) cluster(state year)

(dropped 16 singleton observations)
(MWFE estimator converged in 27 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.

HDFE Linear regression                            Number of obs   =     14,894
Absorbing 4 HDFE groups                           F(   3,     47) =       0.62
Statistics robust to heteroskedasticity           Prob > F        =     0.6055
                                                  R-squared       =     0.5753
                                                  Adj R-squared   =     0.4875
Number of clusters (state)   =         50         Within R-sq.    =     0.0004
Number of clusters (year)    =         48         Root MSE        = 23294.9124

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
   txt_words |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   4179.933   3196.276     1.31   0.197    -2250.145    10610.01
 ptonpselect |  -425.9655   1274.101    -0.33   0.740    -2989.126    2137.195
 nptomselect |  -2831.691   8478.891    -0.33   0.740    -19889.02    14225.64
       _cons |   56250.18   238.7813   235.57   0.000     55769.82    56730.55
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est1 stored)

(dropped 16 singleton observations)
(MWFE estimator converged in 27 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.
warning: missing F statistic; dropped variables due to collinearity or too few c
> lusters

HDFE Linear regression                            Number of obs   =     14,894
Absorbing 4 HDFE groups                           F(   3,     47) =          .
Statistics robust to heteroskedasticity           Prob > F        =          .
                                                  R-squared       =     0.5593
                                                  Adj R-squared   =     0.4682
Number of clusters (state)   =         50         Within R-sq.    =     0.0000
Number of clusters (year)    =         48         Root MSE        =     0.4531

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
txt_wo~s_log |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |  -.0064813   .0499828    -0.13   0.897    -.1070337    .0940712
 ptonpselect |  -.0162006   .0232369    -0.70   0.489    -.0629472     .030546
 nptomselect |  -.0226066    .136415    -0.17   0.869    -.2970382     .251825
       _cons |   10.76995   .0033815  3184.97   0.000     10.76314    10.77675
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est2 stored)

(dropped 10 singleton observations)
(MWFE estimator converged in 27 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.

HDFE Linear regression                            Number of obs   =     14,894
Absorbing 4 HDFE groups                           F(   3,     47) =       1.93
Statistics robust to heteroskedasticity           Prob > F        =     0.1372
                                                  R-squared       =     0.0235
                                                  Adj R-squared   =    -0.1784
Number of clusters (state)   =         50         Within R-sq.    =     0.0003
Number of clusters (year)    =         48         Root MSE        =     0.3704

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_txt_ch~s |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .0407632   .0393244     1.04   0.305    -.0383473    .1198736
 ptonpselect |   .0731709   .0503196     1.45   0.153     -.028059    .1744008
 nptomselect |  -.0614412   .1296384    -0.47   0.638      -.32224    .1993577
       _cons |   .4968349   .0030004   165.59   0.000     .4907988    .5028709
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est3 stored)

(dropped 10 singleton observations)
(MWFE estimator converged in 27 iterations)

HDFE Linear regression                            Number of obs   =     14,894
Absorbing 4 HDFE groups                           F(   3,     47) =       1.01
Statistics robust to heteroskedasticity           Prob > F        =     0.3987
                                                  R-squared       =     0.0251
                                                  Adj R-squared   =    -0.1765
Number of clusters (state)   =         50         Within R-sq.    =     0.0003
Number of clusters (year)    =         48         Root MSE        =     0.3702

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_txt_se~s |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |    .048748   .0415978     1.17   0.247    -.0349359    .1324319
 ptonpselect |   .0576388   .0640272     0.90   0.373    -.0711673    .1864448
 nptomselect |  -.0246423   .1257323    -0.20   0.845    -.2775832    .2282985
       _cons |   .4962383   .0032211   154.06   0.000     .4897583    .5027183
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est4 stored)
In [210]:
%html
esttab, r2 se star(+ 0.10 * 0.05 ** 0.01) drop(   _cons) html
esttab using "1-select/A-output-specs-stars.csv", replace r2 se star(+ 0.10 * 0.05 ** 0.01) drop(   _cons)

(1) (2) (3) (4)
txt_words txt_words_log pct_txt_chars pct_txt_sents

ptomselect 4179.9 -0.00648 0.0408 0.0487
(3196.3) (0.0500) (0.0393) (0.0416)
 
ptonpselect -426.0 -0.0162 0.0732 0.0576
(1274.1) (0.0232) (0.0503) (0.0640)
 
nptomselect -2831.7 -0.0226 -0.0614 -0.0246
(8478.9) (0.136) (0.130) (0.126)

N 14894 14894 14894 14894
R2 0.575 0.559 0.024 0.025

Standard errors in parentheses
+ p < 0.10, * p < 0.05, ** p < 0.01
(output written to 1-select/A-output-specs-stars.csv)

Additional Judge Behaviors

In [211]:
local additional_vars_1 "pct_action_affirm_m pct_txt_words_m pct_toc_unique_m pct_dis_numops pct_cites_pos pct_cites_overruled_m pct_cites_supersede_m"
In [212]:
eststo clear

foreach y in `additional_vars_1' {
    
eststo: reghdfe `y'  ptomselect ptonpselect nptomselect  , absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) ) cluster(state year)

}

(dropped 10 singleton observations)
(MWFE estimator converged in 27 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.

HDFE Linear regression                            Number of obs   =     14,894
Absorbing 4 HDFE groups                           F(   3,     47) =       0.17
Statistics robust to heteroskedasticity           Prob > F        =     0.9161
                                                  R-squared       =     0.0203
                                                  Adj R-squared   =    -0.1823
Number of clusters (state)   =         50         Within R-sq.    =     0.0001
Number of clusters (year)    =         48         Root MSE        =     0.3708

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_action~m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .0202301   .0468674     0.43   0.668     -.074055    .1145152
 ptonpselect |   .0474337   .1256144     0.38   0.707    -.2052699    .3001372
 nptomselect |    -.02142   .0533094    -0.40   0.690    -.1286647    .0858248
       _cons |     .50073   .0040053   125.02   0.000     .4926724    .5087875
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est1 stored)
(dropped 10 singleton observations)
(MWFE estimator converged in 27 iterations)

HDFE Linear regression                            Number of obs   =     14,894
Absorbing 4 HDFE groups                           F(   3,     47) =       1.08
Statistics robust to heteroskedasticity           Prob > F        =     0.3667
                                                  R-squared       =     0.0336
                                                  Adj R-squared   =    -0.1662
Number of clusters (state)   =         50         Within R-sq.    =     0.0008
Number of clusters (year)    =         48         Root MSE        =     0.3687

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_txt_wo~m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .0665471   .0696018     0.96   0.344    -.0734737    .2065679
 ptonpselect |   -.055484   .0410627    -1.35   0.183    -.1380915    .0271235
 nptomselect |  -.1040083   .1291294    -0.81   0.425    -.3637831    .1557665
       _cons |   .4976912   .0046708   106.55   0.000     .4882947    .5070877
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est2 stored)
(dropped 10 singleton observations)
(MWFE estimator converged in 27 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.

HDFE Linear regression                            Number of obs   =     14,894
Absorbing 4 HDFE groups                           F(   3,     47) =       0.54
Statistics robust to heteroskedasticity           Prob > F        =     0.6545
                                                  R-squared       =     0.0357
                                                  Adj R-squared   =    -0.1637
Number of clusters (state)   =         50         Within R-sq.    =     0.0009
Number of clusters (year)    =         48         Root MSE        =     0.3684

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_toc_un~m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .0840689    .080051     1.05   0.299     -.076973    .2451109
 ptonpselect |      .0156    .074453     0.21   0.835      -.13418    .1653801
 nptomselect |  -.1057169   .1583136    -0.67   0.508    -.4242026    .2127689
       _cons |   .4962376   .0058321    85.09   0.000      .484505    .5079702
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est3 stored)
(dropped 10 singleton observations)
(MWFE estimator converged in 27 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.

HDFE Linear regression                            Number of obs   =     14,894
Absorbing 4 HDFE groups                           F(   3,     47) =       0.67
Statistics robust to heteroskedasticity           Prob > F        =     0.5747
                                                  R-squared       =     0.0782
                                                  Adj R-squared   =    -0.1124
Number of clusters (state)   =         50         Within R-sq.    =     0.0003
Number of clusters (year)    =         48         Root MSE        =     0.3559

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_dis_nu~s |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |  -.0321321   .0492375    -0.65   0.517    -.1311851    .0669209
 ptonpselect |   .0848585   .0726278     1.17   0.249    -.0612498    .2309667
 nptomselect |  -.0536337   .0890394    -0.60   0.550     -.232758    .1254906
       _cons |   .5255766   .0033333   157.67   0.000     .5188708    .5322824
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est4 stored)
(dropped 10 singleton observations)
(MWFE estimator converged in 27 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.
warning: missing F statistic; dropped variables due to collinearity or too few c
> lusters

HDFE Linear regression                            Number of obs   =     14,894
Absorbing 4 HDFE groups                           F(   3,     47) =          .
Statistics robust to heteroskedasticity           Prob > F        =          .
                                                  R-squared       =     0.0313
                                                  Adj R-squared   =    -0.1690
Number of clusters (state)   =         50         Within R-sq.    =     0.0004
Number of clusters (year)    =         48         Root MSE        =     0.3690

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_ci~s_pos |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .0399363   .0333594     1.20   0.237    -.0271742    .1070468
 ptonpselect |    .119905   .0142615     8.41   0.000     .0912146    .1485953
 nptomselect |  -.0150285   .1180566    -0.13   0.899    -.2525278    .2224707
       _cons |   .4959351   .0026791   185.11   0.000     .4905455    .5013247
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est5 stored)
(dropped 10 singleton observations)
(MWFE estimator converged in 27 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.
warning: missing F statistic; dropped variables due to collinearity or too few c
> lusters

HDFE Linear regression                            Number of obs   =     14,894
Absorbing 4 HDFE groups                           F(   3,     47) =          .
Statistics robust to heteroskedasticity           Prob > F        =          .
                                                  R-squared       =     0.4210
                                                  Adj R-squared   =     0.3012
Number of clusters (state)   =         50         Within R-sq.    =     0.0004
Number of clusters (year)    =         48         Root MSE        =     0.2701

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_cite~d_m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .0419346   .0388962     1.08   0.286    -.0363144    .1201836
 ptonpselect |  -.0233345   .0079291    -2.94   0.005    -.0392858   -.0073831
 nptomselect |  -.0411979   .0436515    -0.94   0.350    -.1290133    .0466175
       _cons |   .6112453   .0025861   236.36   0.000     .6060427    .6164479
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est6 stored)
(dropped 10 singleton observations)
(MWFE estimator converged in 27 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.

HDFE Linear regression                            Number of obs   =     14,894
Absorbing 4 HDFE groups                           F(   3,     47) =       1.84
Statistics robust to heteroskedasticity           Prob > F        =     0.1537
                                                  R-squared       =     0.2821
                                                  Adj R-squared   =     0.1337
Number of clusters (state)   =         50         Within R-sq.    =     0.0004
Number of clusters (year)    =         48         Root MSE        =     0.3040

                            (Std. Err. adjusted for 48 clusters in state year)
------------------------------------------------------------------------------
             |               Robust
pct_cit~de_m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
  ptomselect |   .0174284   .0360558     0.48   0.631    -.0551066    .0899633
 ptonpselect |    .063192   .0562438     1.12   0.267    -.0499559    .1763399
 nptomselect |  -.0700842   .0337015    -2.08   0.043    -.1378829   -.0022855
       _cons |   .5679185   .0025553   222.25   0.000     .5627779     .573059
------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est7 stored)
In [213]:
%html
esttab, ar2 se star(+ 0.10 * 0.05 ** 0.01) drop(_cons ) html
esttab using "1-select/A-other-variables-stars.csv", replace r2 se star(+ 0.10 * 0.05 ** 0.01) drop(   _cons)

(1) (2) (3) (4) (5) (6) (7)
pct_action_affirm_m pct_txt_words_m pct_toc_unique_m pct_dis_numops pct_cites_pos pct_cites_overruled_m pct_cites_supersede_m

ptomselect 0.0202 0.0665 0.0841 -0.0321 0.0399 0.0419 0.0174
(0.0469) (0.0696) (0.0801) (0.0492) (0.0334) (0.0389) (0.0361)
 
ptonpselect 0.0474 -0.0555 0.0156 0.0849 0.120** -0.0233** 0.0632
(0.126) (0.0411) (0.0745) (0.0726) (0.0143) (0.00793) (0.0562)
 
nptomselect -0.0214 -0.104 -0.106 -0.0536 -0.0150 -0.0412 -0.0701*
(0.0533) (0.129) (0.158) (0.0890) (0.118) (0.0437) (0.0337)

N 14894 14894 14894 14894 14894 14894 14894
adj. R2 -0.182 -0.166 -0.164 -0.112 -0.169 0.301 0.134

Standard errors in parentheses
+ p < 0.10, * p < 0.05, ** p < 0.01
(output written to 1-select/A-other-variables-stars.csv)

Caseload Characteristics

In [214]:
*// form predictions of cites from case characteristics
*// criminal cases have the biggest effect
eststo clear
eststo: reghdfe pct_cites_pos_m pca_* casetype_*_m, noabsorb cluster(state year)
eststo: reghdfe pct_cites_pos_m pca_* casetype_*_m, absorb(stateyear) cluster(state year)
eststo: reghdfe pct_cites_pos_m pca_* casetype_*_m, absorb(stateyear x_decade_start birth_decade statefe##c.(x_decade_start) ) cluster(state year)

(MWFE estimator converged in 1 iterations)

HDFE Linear regression                            Number of obs   =     14,996
Absorbing 1 HDFE group                            F(   9,     47) =      13.36
Statistics robust to heteroskedasticity           Prob > F        =     0.0000
                                                  R-squared       =     0.0093
                                                  Adj R-squared   =     0.0087
Number of clusters (state)   =         50         Within R-sq.    =     0.0093
Number of clusters (year)    =         48         Root MSE        =     0.3398

                             (Std. Err. adjusted for 48 clusters in state year)
-------------------------------------------------------------------------------
              |               Robust
pct_cites_p~m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
--------------+----------------------------------------------------------------
        pca_1 |   .0022232   .0016314     1.36   0.179    -.0010587    .0055051
        pca_2 |   .0033287   .0014954     2.23   0.031     .0003203    .0063371
        pca_3 |    .001474   .0017851     0.83   0.413    -.0021171    .0050651
        pca_4 |  -.0011152   .0020355    -0.55   0.586    -.0052102    .0029797
        pca_5 |  -.0070825   .0018811    -3.77   0.000    -.0108667   -.0032982
casetype_ad~m |    .075877   .0462139     1.64   0.107    -.0170933    .1688474
casetype_ci~m |    .194372   .0325859     5.96   0.000     .1288176    .2599265
casetype_co~m |    .215644   .0429755     5.02   0.000     .1291885    .3020995
casetype_~m_m |   .1924146   .0343226     5.61   0.000     .1233664    .2614628
        _cons |   .3141298   .0281168    11.17   0.000     .2575662    .3706934
-------------------------------------------------------------------------------

Absorbed degrees of freedom:
-----------------------------------------------------+
 Absorbed FE | Categories  - Redundant  = Num. Coefs |
-------------+---------------------------------------|
       _cons |         1           0           1     |
-----------------------------------------------------+
(est1 stored)

(MWFE estimator converged in 1 iterations)
Warning: VCV matrix was non-positive semi-definite; adjustment from Cameron, Gel
> bach & Miller applied.

HDFE Linear regression                            Number of obs   =     14,996
Absorbing 1 HDFE group                            F(   9,     47) =      36.67
Statistics robust to heteroskedasticity           Prob > F        =     0.0000
                                                  R-squared       =     0.0209
                                                  Adj R-squared   =    -0.1700
Number of clusters (state)   =         50         Within R-sq.    =     0.0208
Number of clusters (year)    =         48         Root MSE        =     0.3691

                             (Std. Err. adjusted for 48 clusters in state year)
-------------------------------------------------------------------------------
              |               Robust
pct_cites_p~m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
--------------+----------------------------------------------------------------
        pca_1 |   .0085957   .0045947     1.87   0.068    -.0006476    .0178391
        pca_2 |   .0107307   .0024161     4.44   0.000     .0058701    .0155913
        pca_3 |   .0019295   .0040516     0.48   0.636    -.0062214    .0100804
        pca_4 |  -.0049409   .0030956    -1.60   0.117    -.0111684    .0012865
        pca_5 |  -.0108365   .0026333    -4.12   0.000     -.016134    -.005539
casetype_ad~m |   .1502752   .0668349     2.25   0.029     .0158209    .2847296
casetype_ci~m |   .2990528   .0389912     7.67   0.000     .2206125     .377493
casetype_co~m |   .3235402   .0670548     4.83   0.000     .1886434     .458437
casetype_~m_m |   .4937308   .0557435     8.86   0.000     .3815894    .6058723
        _cons |   .1478354   .0334211     4.42   0.000     .0806008    .2150699
-------------------------------------------------------------------------------

Absorbed degrees of freedom:
-----------------------------------------------------+
 Absorbed FE | Categories  - Redundant  = Num. Coefs |
-------------+---------------------------------------|
 stateyearfe |      2438        2438           0    *|
-----------------------------------------------------+
* = FE nested within cluster; treated as redundant for DoF computation
(est2 stored)

(dropped 10 singleton observations)
(MWFE estimator converged in 27 iterations)

HDFE Linear regression                            Number of obs   =     14,894
Absorbing 4 HDFE groups                           F(   9,     47) =      26.67
Statistics robust to heteroskedasticity           Prob > F        =     0.0000
                                                  R-squared       =     0.0652
                                                  Adj R-squared   =    -0.1286
Number of clusters (state)   =         50         Within R-sq.    =     0.0217
Number of clusters (year)    =         48         Root MSE        =     0.3625

                             (Std. Err. adjusted for 48 clusters in state year)
-------------------------------------------------------------------------------
              |               Robust
pct_cites_p~m |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
--------------+----------------------------------------------------------------
        pca_1 |   .0100379   .0036635     2.74   0.009      .002668    .0174078
        pca_2 |   .0093303   .0024888     3.75   0.000     .0043235    .0143371
        pca_3 |    .002009   .0038718     0.52   0.606    -.0057801    .0097981
        pca_4 |  -.0047703    .002775    -1.72   0.092    -.0103529    .0008123
        pca_5 |  -.0107811   .0026459    -4.07   0.000     -.016104   -.0054582
casetype_ad~m |   .1186596   .0588138     2.02   0.049     .0003415    .2369777
casetype_ci~m |    .275766   .0398633     6.92   0.000     .1955714    .3559605
casetype_co~m |   .3118956   .0640187     4.87   0.000     .1831066    .4406846
casetype_~m_m |   .5001116   .0561626     8.90   0.000      .387127    .6130962
        _cons |   .1589639   .0345359     4.60   0.000     .0894867    .2284412
-------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------------------+
              Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------------------+---------------------------------------|
              stateyearfe |      2428        2428           0    *|
           x_decade_start |         9           0           9     |
             birth_decade |        10           1           9     |
                  statefe |        52          52           0    *|
 statefe#c.x_decade_start |        52           0          52    ?|
------------------------------------------------------------------+
? = number of redundant parameters may be higher
* = FE nested within cluster; treated as redundant for DoF computation
(est3 stored)
In [215]:
%html
esttab, ar2 se star(+ 0.10 * 0.05 ** 0.01) drop(_cons ) html
esttab using "1-select/A-reg-pred.csv", replace r2 se star(+ 0.10 * 0.05 ** 0.01) drop(   _cons)

(1) (2) (3)
pct_cites_pos_m pct_cites_pos_m pct_cites_pos_m

pca_1 0.00222 0.00860+ 0.0100**
(0.00163) (0.00459) (0.00366)
 
pca_2 0.00333* 0.0107** 0.00933**
(0.00150) (0.00242) (0.00249)
 
pca_3 0.00147 0.00193 0.00201
(0.00179) (0.00405) (0.00387)
 
pca_4 -0.00112 -0.00494 -0.00477+
(0.00204) (0.00310) (0.00278)
 
pca_5 -0.00708** -0.0108** -0.0108**
(0.00188) (0.00263) (0.00265)
 
casetype_admin_m 0.0759 0.150* 0.119*
(0.0462) (0.0668) (0.0588)
 
casetype_civ_m 0.194** 0.299** 0.276**
(0.0326) (0.0390) (0.0399)
 
casetype_con_m 0.216** 0.324** 0.312**
(0.0430) (0.0671) (0.0640)
 
casetype_crim_m 0.192** 0.494** 0.500**
(0.0343) (0.0557) (0.0562)

N 14996 14996 14894
adj. R2 0.009 -0.170 -0.129

Standard errors in parentheses
+ p < 0.10, * p < 0.05, ** p < 0.01
(output written to 1-select/A-reg-pred.csv)
In [216]:
binscatter pct_cites_pos_m pred_pct_cites_pos_m   , ytitle("Positive Cites per Case (Rank)") xtitle("Caseload-Predicted Positive Cites per Case")
graph display
graph export 1-select/A-binscatter-cites-pred.pdf,replace



(file 1-select/A-binscatter-cites-pred.pdf written in PDF format)
In [ ]:

Selective Attrition

In [217]:
*// judges hit by the reforms do not selectively extend careers 
eststo clear
eststo: reghdfe career_length pre_quality_pct rule_nptou c.rule_nptou#c.pre_quality_pct rule_ptou c.rule_ptou#c.pre_quality_pct rule_ptonp c.rule_ptonp#c.pre_quality_pct if year == x_year_end, absorb(statefe  x_year_start) cluster(state)

(dropped 15 singleton observations)
(MWFE estimator converged in 12 iterations)

HDFE Linear regression                            Number of obs   =        175
Absorbing 2 HDFE groups                           F(   7,     16) =      26.46
Statistics robust to heteroskedasticity           Prob > F        =     0.0000
                                                  R-squared       =     0.6834
                                                  Adj R-squared   =     0.4702
                                                  Within R-sq.    =     0.4365
Number of clusters (state)   =         17         Root MSE        =     5.8016

                                  (Std. Err. adjusted for 17 clusters in state)
-------------------------------------------------------------------------------
              |               Robust
career_length |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
--------------+----------------------------------------------------------------
pre_quality~t |   -1.46753   3.791748    -0.39   0.704    -9.505676    6.570617
   rule_nptou |   9.002786   3.197121     2.82   0.012     2.225192    15.78038
              |
 c.rule_nptou#|
           c. |
pre_quality~t |  -4.693301   6.719963    -0.70   0.495    -18.93899    9.552384
              |
    rule_ptou |   19.17744   5.368854     3.57   0.003     7.795979     30.5589
              |
  c.rule_ptou#|
           c. |
pre_quality~t |  -4.324561   7.266235    -0.60   0.560    -19.72829    11.07917
              |
   rule_ptonp |   7.025473   2.395086     2.93   0.010     1.948118    12.10283
              |
 c.rule_ptonp#|
           c. |
pre_quality~t |  -.4644519   4.125153    -0.11   0.912    -9.209385    8.280481
              |
        _cons |   8.687084   2.048024     4.24   0.001     4.345466     13.0287
-------------------------------------------------------------------------------

Absorbed degrees of freedom:
------------------------------------------------------+
  Absorbed FE | Categories  - Redundant  = Num. Coefs |
--------------+---------------------------------------|
      statefe |        18          18           0    *|
 x_year_start |        46           0          46     |
------------------------------------------------------+
* = FE nested within cluster; treated as redundant for DoF computation
(est1 stored)
In [218]:
%html
esttab, ar2 se star(+ 0.10 * 0.05 ** 0.01) drop(_cons ) html
esttab using "1-select/A-selectve-attrition.csv", replace r2 se star(+ 0.10 * 0.05 ** 0.01) drop(   _cons)

(1)
career_length

pre_quality_pct -1.468
(3.792)
 
rule_nptou 9.003*
(3.197)
 
c.rule_nptou#c.pre_quality_pct -4.693
(6.720)
 
rule_ptou 19.18**
(5.369)
 
c.rule_ptou#c.pre_quality_pct -4.325
(7.266)
 
rule_ptonp 7.025**
(2.395)
 
c.rule_ptonp#c.pre_quality_pct -0.464
(4.125)

N 175
adj. R2 0.470

Standard errors in parentheses
+ p < 0.10, * p < 0.05, ** p < 0.01
(output written to 1-select/A-selectve-attrition.csv)

Additional Robustness Checks

Event Study Figure

In [219]:
gen ptom_time = x_year_start - ruleyear_ptou
(12,284 missing values generated)
In [220]:
gen ptonp_time = x_year_start - ruleyear_ptonp
(13,389 missing values generated)
In [221]:
binscatter pct_cites_outstate ptom_time if ptom_time >= -8 & ptom_time <= 10 & ptom_time != 1, nquantiles(16) rd(1)  ytitle("Out-of-State Citations (Rank)") xtitle("Starting Year, Relative to Partisan-to-Merit Reform") xsize(7)
graph display
graph export 1-select/A-event-study-ptom.pdf, replace
warning: nquantiles(16) was specified, but only 14 were generated. see help file
>  under nquantiles() for explanation.