------------------------------------------------------------------------------------------------- name: log: /Users/djaeger/Dropbox/research/SixteenAndPregnant/6docs/manuscript/IREE/replication > //iree-jjk-replication2.log log type: text opened on: 31 Dec 2018, 09:26:07 . **************************************************************************** . * . * Replication File . * . * "Tweet Sixteen and Pregnant: Missing Links in the Causal Chain from . * Reality TV to Fertility" . * David A. Jaeger, Theodore Joyce, and Robert Kaestner . * . * This paper replicates and extends the results of Kearney, Melissa S. and . * Phillip B. Levine, "Media Influences on Social Outcomes: The Impact of . * MTV's 16 and Pregnant on Teen Childbearing," American Economic Review . * 105(12):3597-3632, December 2015 (henceforth KL). . * . * Resubmitted to Interational Re-Views in Empirical Economics, December 2018 . * . * . * Data used:] . * The data used come partially from the replication files made available by . * KL on the American Economic Review website, which can be downloaded at . * https://www.aeaweb.org/aer/data/10512/20140012_data.zip . * . * Tables 1 and 2 use national-level data from Twitter from the file . * topsy-trend-daily.dta, which is included in KL's replication file above. . * . * Table 3 uses data from * the files topsy-state-daily.dta and . * topsy-state.dta, which are also included in the downloadable zip file. . * . * Table 4 uses national-level data from Google from the file . * googletrends.dta, which was obtained directly from Phil Levine on . * 28 August 2016. This file is referenced in KL's replication files on . * the AER website, but not included in the downloadable zip file. . * . * . **************************************************************************** . . set more off . . clear . . ************ . * Data: Figure 1: National Twitter Results . * Data file obtained from https://www.aeaweb.org/aer/data/10512/20140012_data.zip . . use "$work/data/topsy-trend-daily.dta" . . tsset trend time variable: trend, 1 to 1461 delta: 1 unit . sort trend . . **** . * tweet rates and log tweet rates . * 16p = 16 and Pregnant . * bc = birth control . * abort = abortion . . gen twtrate16p = 1000000*tweets16p/totaltweets (6 missing values generated) . gen twtratebc = 1000000*tweetsbc/totaltweets (6 missing values generated) . gen twtrateabort = 1000000*tweetsabort/totaltweets (6 missing values generated) . . gen lntwtrate16p = log(twtrate16p) (139 missing values generated) . gen lntwtratebc = log(twtratebc) (6 missing values generated) . gen lntwtrateabort = log(twtrateabort) (6 missing values generated) . . label var lntwtrate16p "16 and Pregnant" . label var lntwtratebc "Birth Control" . label var lntwtrateabor "Abortion" . . **** . . **** . * Generate dummy variables for different peridos of show . gen date2 =date(date,"MDY") . label var date2 " " . tsset date2, daily time variable: date2, 01jan2009 to 31dec2012 delta: 1 day . . gen year2012=(date2>td(31dec2011)) . gen year2011=(date2>td(31dec2010) & date2td(31dec2009) & date2td(31dec2008) & date2 > graph twoway > (tsline lntwtratebc if year2010==1, > lwidth(thin) lcolor(blue) scheme("s1mono") > plotregion(lstyle(none)) > xtitle("") > ytitle("Log(Tweet Rates)", size(small)) > ) > > || > > (tsline lntwtrate16p if year2010==1 , > lcolor(red) lwidth(thin) > tline(16feb2010 > 23feb2010 > 02mar2010 > 09mar2010 > 16mar2010 > 23mar2010 > 30mar2010 > 06apr2010 > 13apr2010 > 20apr2010 > 26oct2010 > 02nov2010 > 09nov2010 > 16nov2010 > 23nov2010 > 30nov2010 > 06dec2010 > 14dec2010 > 21dec2010 > 28dec2010, > lcolor(black) lwidth(vthin)) > tlabel(01feb2010 01jun2010 01sep2010 01dec2010, format(%tdmd) labsize(small)) > ylabel(, labsize(small))), > title("{it: Panel A: Birth Control}", size(small)) > legend(cols(2) region(lstyle(none)) size(small) > lab(1 "Birth Control") lab(2 "{it:16 and Pregnant}") ) > name(Fig1PanelA, replace); . /* Figure 1, Panel B */ > > graph twoway > (tsline lntwtrateabort if year2010==1, > lwidth(thin) lcolor(green) scheme("s1mono") > plotregion(lstyle(none)) > plotregion(lstyle(none)) > xtitle("") > ytitle("Log(Tweet Rates)", size(small)) > ) > > || > > > (tsline lntwtrate16p if year2010==1 , > lcolor(red) lwidth(thin) > > tline(16feb2010 > 23feb2010 > 02mar2010 > 09mar2010 > 16mar2010 > 23mar2010 > 30mar2010 > 06apr2010 > 13apr2010 > 20apr2010 > 26oct2010 > 02nov2010 > 09nov2010 > 16nov2010 > 23nov2010 > 30nov2010 > 06dec2010 > 14dec2010 > 21dec2010 > 28dec2010, > lcolor(black) lwidth(vthin)) > tlabel(01feb2010 01jun2010 01sep2010 01dec2010, format(%tdmd) labsize(small)) > ylabel(, labsize(small))), > title("{it:Panel B: Abortion}", size(small)) > legend(cols(2) region(lstyle(none)) size(small) > lab(1 "Abortion") lab(2 "{it:16 and Pregnant}") ) > name(Fig1PanelB, replace); . graph combine > Fig1PanelA > Fig1PanelB, > col(1) > ysize(11) > xsize(8.5) > xcommon > scheme("s1mono") > title("{stSerif:{bf:Figure 1}}" "{stSerif:Twitter Activity for {it:16 and Pregn > ant}, Birth Control, and Abortion}" > "{stSerif:Season 2 (2010)}", size(2.25) ) > graphregion(margin(11.76 11.76 11.76 11.76)) > note("{stSerif:{it:Notes:} These graphs plot the log tweet rate for {it:16 and > Pregnant}, birth control, and abortion}" > "{stSerif:for the second season (2010) of {it:16 and Pregnant}. The verti > cal lines represent the days of}" > "{stSerif:broadcasts of {it:16 and Pregnant}. Note that the vertical scale is > different in each panel.}", size(2) ) > ; . graph export "$work/Figure1.pdf", replace; (file ~/Dropbox/research/SixteenAndPregnant/6docs/manuscript/IREE/replication//Figure1.pdf writte > n in PDF format) . #delimit cr delimiter now cr . . clear . . ************ . . . ************ . * Data: Figure 2, nationa Twitter results . * Data file obtained from https://www.aeaweb.org/aer/data/10512/20140012_data.zip . * Shading from the corrected "in season" dates referenced under "Broadcast Date Corrections" be > low . . use "$work/data/topsy-trend-daily.dta" . . **** . * Generate dummy variables for different peridos of show . gen date2 =date(date,"MDY") . label var date2 " " . tsset date2, daily time variable: date2, 01jan2009 to 31dec2012 delta: 1 day . . sort trend . tsset trend time variable: trend, 1 to 1461 delta: 1 unit . . **** . * Rescale totaltweets variable (to millions) . . gen totaltweetsmillion=totaltweets/1000000 . . . *** . * Calculate average weight for different "in season" periods . * Uses corrected "in season" dates described below . * For discussion in text . gen seasonnum=0 . replace seasonnum=1 if trend >= 162 & trend <=211 (50 real changes made) . replace seasonnum=2 if trend >= 412 & trend <=475 (64 real changes made) . replace seasonnum=3 if trend >= 664 & trend <=734 (71 real changes made) . replace seasonnum=4 if trend >= 839 & trend <=909 (71 real changes made) . replace seasonnum=5 if trend >= 1182 & trend <= 1252 (71 real changes made) . . forvalues s=1/5 { 2. summarize totaltweetsmillion if seasonnum==`s' 3. } Variable | Obs Mean Std. Dev. Min Max -------------+--------------------------------------------------------- totaltweet~n | 50 5.0768 .9832608 2.88 6.75 Variable | Obs Mean Std. Dev. Min Max -------------+--------------------------------------------------------- totaltweet~n | 64 10.47984 .6440928 8.93 11.65 Variable | Obs Mean Std. Dev. Min Max -------------+--------------------------------------------------------- totaltweet~n | 71 12.56845 1.295388 9.54 15.28 Variable | Obs Mean Std. Dev. Min Max -------------+--------------------------------------------------------- totaltweet~n | 71 23.87761 2.524181 19.47 32.04 Variable | Obs Mean Std. Dev. Min Max -------------+--------------------------------------------------------- totaltweet~n | 71 60.65211 4.008044 52.05 67.59 . . . *** . * Graph . . #delimit ; delimiter now ; . graph twoway > > (function y=80, recast(area) bfcolor(gs12) blcolor(gs12) range(162 211) ) || > (function y=80, recast(area) bfcolor(gs12) blcolor(gs12) range(412 475) ) || > (function y=80, recast(area) bfcolor(gs12) blcolor(gs12) range(664 734) ) || > (function y=80, recast(area) bfcolor(gs12) blcolor(gs12) range(839 909) ) || > (function y=80, recast(area) bfcolor(gs12) blcolor(gs12) range(1182 1252) ) || > > (line totaltweetsmillion trend, > lwidth(thin) lcolor(black) scheme("s1mono") > plotregion(lstyle(none)) > xtitle("") > ytitle("Total Tweets (millions)", size(vsmall)) > xlabel(1 "Jan 2009" 366 "Jan 2010" 731 "Jan 2011" 1096 "Jan 2012" 1462 "Jan 2013", labsize(vsma > ll)) > ylabel(, labsize(vsmall)) > ) > > , legend(off) > name(Fig2, replace) > ; . graph combine > Fig2, > col(1) > ysize(8.5) > xsize(11) > > scheme("s1mono") > title("{stSerif:{bf:Figure 2}}" "{stSerif:Total Number of Tweets}" > "{stSerif:2009-2012}", size(2.25) ) > graphregion(margin(11.76 11.76 11.76 11.76)) > note("{stSerif:{it:Notes:} This graph shows the total number of daily tweets f > rom 2008 to 2012, which KL use as weights in }" > "{stSerif:the national-level Twitter regressions shown in Table 2. The sha > ded areas indicate ''in season'' periods}" > "{stSerif:for {it:16 and Pregnant}, using the corrected dates as discu > ssed in the text.}", size(2) ) > ; . graph export "$work/Figure2.pdf", replace; (file ~/Dropbox/research/SixteenAndPregnant/6docs/manuscript/IREE/replication//Figure2.pdf writte > n in PDF format) . #delimit cr delimiter now cr . . clear . . ************ . . . ************ . * Data: Tables 1 and 2, national Twitter results . * Data file obtained from https://www.aeaweb.org/aer/data/10512/20140012_data.zip . . use "$work/data/topsy-trend-daily.dta" . . tsset trend time variable: trend, 1 to 1461 delta: 1 unit . sort trend . . **** . * tweet rates and log tweet rates . * 16p = 16 and Pregnant . * bc = birth control . * abort = abortion . . gen twtrate16p = 1000000*tweets16p/totaltweets (6 missing values generated) . gen twtratebc = 1000000*tweetsbc/totaltweets (6 missing values generated) . gen twtrateabort = 1000000*tweetsabort/totaltweets (6 missing values generated) . . gen lntwtrate16p = log(twtrate16p) (139 missing values generated) . gen lntwtratebc = log(twtratebc) (6 missing values generated) . gen lntwtrateabort = log(twtrateabort) (6 missing values generated) . . **** . . . **** . * Variable labels for KL's "in season" and "day show was broadcast" variables . * . * showon1 is an indicator for period in which show was airing ("in season") . * showon2 is an indicator for the day on which the show was broadcast . * . * rename KL's variables . rename showon1 inseason . rename showon2 broadcastday . . **** . . **** . * Trend quadratic . gen trend2 = trend^2 . **** . . ************ . . . . . ************ . * Broadcast date corrections . * . * As noted in the text, 11 of the 53 days on which 16P was broadcast were . * miscoded by one day, implying that 11 lagged variables were also miscoded . * by one day. . * . * We corrected the dates of broadcast using the following sources: . * . * 16 and Pregnant . * 1st season is Jun 2009 - Jul 2009, . * 2nd season is Feb 2010 - Dec 2010, . * 3rd season is Apr 2011 - Jun 2011. . * . * http://www.mtv.com/shows/16-and-pregnant/episode-guide (official MTV site) . * http://www.tvguide.com/tvshows/16-pregnant/episodes-season-1/304110/ . * http://www.tvguide.com/tvshows/16-pregnant/episodes-season-2/304110/ . * http://www.tvguide.com/tvshows/16-pregnant/episodes-season-3/304110/ . * . * . * Teen Mom and Teen Mom 2 . * Teen Mom 1st season is Dec 2009 - Feb 2010 . * Teen Mom 2 1st seaons is Jan 2011 - Mar 2011 . * . * http://www.mtv.com/shows/teen-mom/episode-guide (teen mom season 1 not listed . * by KL Dec 2009 - Feb 2010) . * http://www.mtv.com/shows/teen-mom-2/episode-guide . * . * We cannot find 16P, Teen Mom, or Teen Mom2 broadcast in Feb 2010 - Apr 2010 . * . * These are KL's "in season" dates, which which are incorrect: . * period 1 = 1/1/2009 - 6/10/2009: off . * period 2 = 6/11/2009 - 7/31/2009: on . * period 3 = 8/1/2009 - 2/15/2010: off . * period 4 = 2/16/2010 - 4/20/2010: on . * period 5 = 4/21/2010 - 10/25/2010: off . * period 6 = 10/26/2010 - 1/5/2011: on . * period 7 = 1/6/2011 - 4/18/2011: off . * period 8 = 4/19/2011 - 6/28/2011: on . * period 9 = 6/29/2011 - 3/20/2012: off . * period 10 = 3/21/2012 - 6/6/2012: on . * period 11 = 6/7/2012 - 12/31/2012: off . * . * These are the correct "in season" dates according to the TV guide sites above . * period 1 = 1/1/2009 - 6/10/2009: off . * period 2 = 6/11/2009 - 7/30/2009: on . * period 3 = 7/31/2009 - 2/15/2010: off . * period 4 = 2/16/2010 - 4/20/2010: on . * period 5 = 4/21/2010 - 10/25/2010: off . * period 6 = 10/26/2010 - 1/4/2011: on . * period 7 = 1/5/2011 - 4/18/2011: off . * period 8 = 4/19/2011 - 6/28/2011: on . * period 9 = 6/29/2011 - 3/26/2012: off . * period 10 = 3/27/2012 - 6/5/2012: on . * period 11 = 6/6/2012 - 12/31/2012: off . * . ************ . . **** . * Create corrected "in season" and "broadcast day" variables . * r_ prefix variables indicate using revised dates . gen r_broadcastday=broadcastday . gen r_inseason=inseason . . . replace r_broadcastday=1 if date=="7/23/2009" (1 real change made) . replace r_broadcastday=0 if date=="7/24/2009" (1 real change made) . . replace r_broadcastday=1 if date=="7/30/2009" (1 real change made) . replace r_broadcastday=0 if date=="7/31/2009" (1 real change made) . replace r_inseason =0 if date=="7/31/2009" (1 real change made) . . replace r_broadcastday=1 if date=="1/4/2011" (1 real change made) . replace r_broadcastday=0 if date=="1/5/2011" (1 real change made) . replace r_inseason =0 if date=="1/5/2011" (1 real change made) . . replace r_broadcastday=0 if date=="3/21/2012" (1 real change made) . . replace r_broadcastday=1 if date=="4/24/2012" (1 real change made) . replace r_broadcastday=0 if date=="4/25/2012" (1 real change made) . . replace r_broadcastday=1 if date=="5/1/2012" (1 real change made) . replace r_broadcastday=0 if date=="5/2/2012" (1 real change made) . . replace r_broadcastday=1 if date=="5/8/2012" (1 real change made) . replace r_broadcastday=0 if date=="5/9/2012" (1 real change made) . . replace r_broadcastday=1 if date=="5/15/2012" (1 real change made) . replace r_broadcastday=0 if date=="5/16/2012" (1 real change made) . . replace r_broadcastday=1 if date=="5/22/2012" (1 real change made) . replace r_broadcastday=0 if date=="5/23/2012" (1 real change made) . . replace r_broadcastday=1 if date=="5/29/2012" (1 real change made) . replace r_broadcastday=0 if date=="5/30/2012" (1 real change made) . replace r_broadcastday=0 if date=="5/31/2012" (0 real changes made) . > . replace r_broadcastday=1 if date=="6/5/2012" (1 real change made) . replace r_broadcastday=0 if date=="6/6/2012" (1 real change made) . replace r_inseason =0 if date=="6/6/2012" (1 real change made) . . ************ . . ************ . * Create lagged broadcast days . * Note: doing this here to have 1/1/2009 as zero rather than missing . gen L_broadcastday=L.broadcastday (1 missing value generated) . replace L_broadcastday=0 if L_broadcastday==. (1 real change made) . . gen L_r_broadcastday=L.r_broadcastday (1 missing value generated) . replace L_r_broadcastday=0 if L_r_broadcastday==. (1 real change made) . . ************ . . ************ . * Create indicators for various broadcast and non-broadcast days . . *** . * Create pre-16 and Pregnant indicator . * numdate==18059 is 11 June 2009, the first day of 16P broadcast . gen numdate=date(date,"MDY") . gen pre16p=numdate<18059 . . *** . * Create "out of season" indicator (0 in pre-16P period) . gen r_outofseason=1-r_inseason . replace r_outofseason=0 if pre16p==1 | L_r_broadcastday==1 (166 real changes made) . . *** . . ************ . . . . ************ . * Replicate KL's results for our Table 1, columns 2 and 5 . * and our Table 2, columns 1 and 4 . * . * Note that the indicator for the day after the show is on is calculated . * incorrectly because KL first drop observations that are not "in season" . * (inseason==0) and then create the lag indicator for the day after a . * broadcast of 16P using the _n-1 construction: broadcastday[_n-1] == 1 . * . * See footnotes 11 and 13 . . * Preserve all data . preserve . . . * keep only "in season" observations . keep if inseason (1,125 observations deleted) . . **** . * Indicator for day after show: Now that because keeping only "in season" observations . * means that the periods are non-contiguous, creating the lag at this point is incorrect . gen lagbroadcastday = broadcastday[_n-1] == 1 . . . **** . * Replicate KL's (incorrect) results . . * Replicate KL Table 3, column 4 (corresponds to our Table 1, column 2 below) . reg lntwtratebc broadcastday lagbroadcastday trend trend2 [weight=totaltweets], robust (analytic weights assumed) (sum of wgt is 8.2331e+09) Linear regression Number of obs = 336 F(4, 331) = 8.98 Prob > F = 0.0000 R-squared = 0.1013 Root MSE = .31194 --------------------------------------------------------------------------------- | Robust lntwtratebc | Coef. Std. Err. t P>|t| [95% Conf. Interval] ----------------+---------------------------------------------------------------- broadcastday | .1204994 .0468217 2.57 0.010 .0283938 .212605 lagbroadcastday | .2286796 .0578758 3.95 0.000 .1148288 .3425304 trend | .0007117 .000261 2.73 0.007 .0001984 .0012251 trend2 | -3.44e-07 1.67e-07 -2.07 0.040 -6.72e-07 -1.66e-08 _cons | 3.316807 .0937121 35.39 0.000 3.132461 3.501154 --------------------------------------------------------------------------------- . . * Replicate KL Table 3, column 5 (corresponds to our Table 1, column 5 below) . reg lntwtrateabort broadcastday lagbroadcastday trend trend2 [weight=totaltweets], robust (analytic weights assumed) (sum of wgt is 8.2331e+09) Linear regression Number of obs = 336 F(4, 331) = 30.43 Prob > F = 0.0000 R-squared = 0.2856 Root MSE = .30776 --------------------------------------------------------------------------------- | Robust lntwtrateabort | Coef. Std. Err. t P>|t| [95% Conf. Interval] ----------------+---------------------------------------------------------------- broadcastday | .1421102 .0362498 3.92 0.000 .0708011 .2134193 lagbroadcastday | .2116535 .0458956 4.61 0.000 .1213698 .3019373 trend | -.00027 .0002721 -0.99 0.322 -.0008052 .0002652 trend2 | -1.84e-07 1.62e-07 -1.14 0.255 -5.02e-07 1.34e-07 _cons | 4.786382 .1130306 42.35 0.000 4.564033 5.008731 --------------------------------------------------------------------------------- . . * Replicate KL Table 4, panel A, column 4 (corresponds to our Table 2, column 1 below) . reg lntwtratebc lntwtrate16p trend trend2 [weight=totaltweets], robust (analytic weights assumed) (sum of wgt is 8.2331e+09) Linear regression Number of obs = 336 F(3, 332) = 7.67 Prob > F = 0.0001 R-squared = 0.0475 Root MSE = .32066 ------------------------------------------------------------------------------ | Robust lntwtratebc | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- lntwtrate16p | .0774588 .0338101 2.29 0.023 .0109497 .143968 trend | .0007431 .0002643 2.81 0.005 .0002232 .001263 trend2 | -3.58e-07 1.70e-07 -2.10 0.036 -6.93e-07 -2.27e-08 _cons | 3.095717 .141177 21.93 0.000 2.818003 3.373432 ------------------------------------------------------------------------------ . . * Replicate KL Table 4, panel A, column 5 (corresponds to our Table 1, column 5 below) . reg lntwtrateabort lntwtrate16p trend trend2 [weight=totaltweets], robust (analytic weights assumed) (sum of wgt is 8.2331e+09) Linear regression Number of obs = 336 F(3, 332) = 30.99 Prob > F = 0.0000 R-squared = 0.2377 Root MSE = .31744 ------------------------------------------------------------------------------ | Robust lntwtratea~t | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- lntwtrate16p | .063525 .0245646 2.59 0.010 .0152031 .1118469 trend | -.0002374 .0002786 -0.85 0.395 -.0007854 .0003105 trend2 | -2.00e-07 1.67e-07 -1.20 0.233 -5.29e-07 1.29e-07 _cons | 4.613925 .14142 32.63 0.000 4.335732 4.892117 ------------------------------------------------------------------------------ . . **** . . * Restore full data set . restore . . *********** . . . . . *********** . * Table 1 Results: Association between 16P broadcast days and Twitter activity . . **** . * Column 1: Association betweeen tweet rate and broadcast of 16P, all dates . newey lntwtrate16p r_broadcastday L_r_broadcastday trend trend2 pre16p r_outofseason, lag(1) fo > rce Regression with Newey-West standard errors Number of obs = 1,322 maximum lag: 1 F( 6, 1315) = 114.79 Prob > F = 0.0000 ---------------------------------------------------------------------------------- | Newey-West lntwtrate16p | Coef. Std. Err. t P>|t| [95% Conf. Interval] -----------------+---------------------------------------------------------------- r_broadcastday | -.2327438 .1120882 -2.08 0.038 -.4526349 -.0128526 L_r_broadcastday | 1.25323 .1483016 8.45 0.000 .9622963 1.544164 trend | .0041285 .0004964 8.32 0.000 .0031546 .0051025 trend2 | -1.65e-06 2.71e-07 -6.09 0.000 -2.18e-06 -1.12e-06 pre16p | -1.016173 .2840213 -3.58 0.000 -1.573358 -.4589887 r_outofseason | -1.267551 .0803715 -15.77 0.000 -1.425222 -1.109881 _cons | 1.317097 .2235364 5.89 0.000 .8785704 1.755624 ---------------------------------------------------------------------------------- . lincom r_broadcastday - pre16p ( 1) r_broadcastday - pre16p = 0 ------------------------------------------------------------------------------ lntwtrate16p | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- (1) | .7834294 .285623 2.74 0.006 .2231028 1.343756 ------------------------------------------------------------------------------ . lincom L_r_broadcastday - pre16p ( 1) L_r_broadcastday - pre16p = 0 ------------------------------------------------------------------------------ lntwtrate16p | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- (1) | 2.269403 .3125495 7.26 0.000 1.656253 2.882553 ------------------------------------------------------------------------------ . lincom r_broadcastday - r_outofseason ( 1) r_broadcastday - r_outofseason = 0 ------------------------------------------------------------------------------ lntwtrate16p | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- (1) | 1.034807 .1037637 9.97 0.000 .8312469 1.238368 ------------------------------------------------------------------------------ . lincom L_r_broadcastday - r_outofseason ( 1) L_r_broadcastday - r_outofseason = 0 ------------------------------------------------------------------------------ lntwtrate16p | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- (1) | 2.520781 .1457105 17.30 0.000 2.234931 2.806632 ------------------------------------------------------------------------------ . . **** . * Column 2: Birth control tweets, Replicate KL Table 3, column 2, with uncorrected dates but c > orrect lag variables, "in season" only . reg lntwtratebc broadcastday L_broadcastday trend trend2 [weight=totaltweets] if inseason==1, r > obust (analytic weights assumed) (sum of wgt is 8.2331e+09) Linear regression Number of obs = 336 F(4, 331) = 8.72 Prob > F = 0.0000 R-squared = 0.1020 Root MSE = .31182 -------------------------------------------------------------------------------- | Robust lntwtratebc | Coef. Std. Err. t P>|t| [95% Conf. Interval] ---------------+---------------------------------------------------------------- broadcastday | .1400051 .0485596 2.88 0.004 .0444808 .2355294 L_broadcastday | .2403974 .0617384 3.89 0.000 .1189483 .3618466 trend | .0007244 .000261 2.78 0.006 .000211 .0012378 trend2 | -3.52e-07 1.67e-07 -2.12 0.035 -6.80e-07 -2.47e-08 _cons | 3.310755 .0939079 35.26 0.000 3.126024 3.495487 -------------------------------------------------------------------------------- . . **** . * Column 3: Birth control tweets, Replicate KL Table 3, column 2, with corrected dates and cor > rect lag variables, "in season" only . reg lntwtratebc r_broadcastday L_r_broadcastday trend trend2 [weight=totaltweets] if r_inseason > ==1, robust (analytic weights assumed) (sum of wgt is 8.1501e+09) Linear regression Number of obs = 333 F(4, 328) = 6.84 Prob > F = 0.0000 R-squared = 0.0678 Root MSE = .31888 ---------------------------------------------------------------------------------- | Robust lntwtratebc | Coef. Std. Err. t P>|t| [95% Conf. Interval] -----------------+---------------------------------------------------------------- r_broadcastday | .0419825 .0568247 0.74 0.461 -.0698043 .1537694 L_r_broadcastday | .2023654 .0604267 3.35 0.001 .0834925 .3212383 trend | .0007224 .0002665 2.71 0.007 .0001981 .0012467 trend2 | -3.52e-07 1.71e-07 -2.05 0.041 -6.88e-07 -1.50e-08 _cons | 3.334462 .0941237 35.43 0.000 3.1493 3.519625 ---------------------------------------------------------------------------------- . . **** . * Column 4: Birth control tweets, Replicate KL Table 3, column 2, with corrected dates and cor > rect lag variables, "in season" only, no weights . reg lntwtratebc r_broadcastday L_r_broadcastday trend trend2 if r_inseason==1, robust Linear regression Number of obs = 333 F(4, 328) = 24.73 Prob > F = 0.0000 R-squared = 0.1579 Root MSE = .3472 ---------------------------------------------------------------------------------- | Robust lntwtratebc | Coef. Std. Err. t P>|t| [95% Conf. Interval] -----------------+---------------------------------------------------------------- r_broadcastday | .0369539 .0448049 0.82 0.410 -.0511872 .1250951 L_r_broadcastday | .1546511 .0508652 3.04 0.003 .0545879 .2547143 trend | .0015647 .0002285 6.85 0.000 .0011152 .0020143 trend2 | -8.93e-07 1.57e-07 -5.70 0.000 -1.20e-06 -5.85e-07 _cons | 3.064519 .0687864 44.55 0.000 2.929201 3.199837 ---------------------------------------------------------------------------------- . . . **** . * Column 5: Birth control tweets, Corrected dates and correct lag variables, all dates . reg lntwtratebc r_broadcastday L_r_broadcastday trend trend2 pre16p r_outofseason [weight=total > tweets] , robust (analytic weights assumed) (sum of wgt is 3.6712e+10) Linear regression Number of obs = 1,455 F(6, 1448) = 21.41 Prob > F = 0.0000 R-squared = 0.0593 Root MSE = .49035 ---------------------------------------------------------------------------------- | Robust lntwtratebc | Coef. Std. Err. t P>|t| [95% Conf. Interval] -----------------+---------------------------------------------------------------- r_broadcastday | .0441949 .0568734 0.78 0.437 -.0673683 .155758 L_r_broadcastday | .1971578 .0560632 3.52 0.000 .0871841 .3071316 trend | .001108 .0002523 4.39 0.000 .000613 .001603 trend2 | -4.71e-07 1.50e-07 -3.13 0.002 -7.66e-07 -1.76e-07 pre16p | .4136654 .0785277 5.27 0.000 .2596253 .5677056 r_outofseason | .2067908 .0348345 5.94 0.000 .1384594 .2751222 _cons | 3.078554 .0952636 32.32 0.000 2.891685 3.265424 ---------------------------------------------------------------------------------- . lincom r_broadcastday - pre16p ( 1) r_broadcastday - pre16p = 0 ------------------------------------------------------------------------------ lntwtratebc | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- (1) | -.3694706 .0897384 -4.12 0.000 -.5455017 -.1934394 ------------------------------------------------------------------------------ . lincom L_r_broadcastday - pre16p ( 1) L_r_broadcastday - pre16p = 0 ------------------------------------------------------------------------------ lntwtratebc | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- (1) | -.2165076 .0893401 -2.42 0.015 -.3917574 -.0412578 ------------------------------------------------------------------------------ . lincom r_broadcastday - r_outofseason ( 1) r_broadcastday - r_outofseason = 0 ------------------------------------------------------------------------------ lntwtratebc | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- (1) | -.1625959 .0552296 -2.94 0.003 -.2709345 -.0542574 ------------------------------------------------------------------------------ . lincom L_r_broadcastday - r_outofseason ( 1) L_r_broadcastday - r_outofseason = 0 ------------------------------------------------------------------------------ lntwtratebc | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- (1) | -.009633 .0543375 -0.18 0.859 -.1162216 .0969557 ------------------------------------------------------------------------------ . . **** . * Column 6: Birth control tweets, Corrected dates and correct lag variables, all dates . newey lntwtratebc r_broadcastday L_r_broadcastday trend trend2 pre16p r_outofseason, lag(1) for > ce Regression with Newey-West standard errors Number of obs = 1,455 maximum lag: 1 F( 6, 1448) = 17.49 Prob > F = 0.0000 ---------------------------------------------------------------------------------- | Newey-West lntwtratebc | Coef. Std. Err. t P>|t| [95% Conf. Interval] -----------------+---------------------------------------------------------------- r_broadcastday | .0401584 .0449102 0.89 0.371 -.0479376 .1282545 L_r_broadcastday | .1576786 .0511647 3.08 0.002 .0573139 .2580434 trend | .0008131 .0001739 4.68 0.000 .000472 .0011541 trend2 | -3.08e-07 1.09e-07 -2.83 0.005 -5.22e-07 -9.45e-08 pre16p | .3089828 .0649333 4.76 0.000 .1816093 .4363562 r_outofseason | .1366606 .0342376 3.99 0.000 .0695 .2038212 _cons | 3.230482 .0659455 48.99 0.000 3.101123 3.359841 ---------------------------------------------------------------------------------- . lincom r_broadcastday - pre16p ( 1) r_broadcastday - pre16p = 0 ------------------------------------------------------------------------------ lntwtratebc | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- (1) | -.2688243 .0702537 -3.83 0.000 -.4066343 -.1310144 ------------------------------------------------------------------------------ . lincom L_r_broadcastday - pre16p ( 1) L_r_broadcastday - pre16p = 0 ------------------------------------------------------------------------------ lntwtratebc | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- (1) | -.1513041 .0769126 -1.97 0.049 -.3021762 -.0004321 ------------------------------------------------------------------------------ . lincom r_broadcastday - r_outofseason ( 1) r_broadcastday - r_outofseason = 0 ------------------------------------------------------------------------------ lntwtratebc | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- (1) | -.0965022 .0414962 -2.33 0.020 -.1779012 -.0151031 ------------------------------------------------------------------------------ . lincom L_r_broadcastday - r_outofseason ( 1) L_r_broadcastday - r_outofseason = 0 ------------------------------------------------------------------------------ lntwtratebc | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- (1) | .021018 .0465835 0.45 0.652 -.0703604 .1123965 ------------------------------------------------------------------------------ . . **** . * Column 7: Abortion tweets, Replicate KL Table 2, column 5 with uncorrected dates but correct > lag variables,"in season" only . reg lntwtrateabort broadcastday L_broadcastday trend trend2 [weight=totaltweets] if inseason==1 > , robust (analytic weights assumed) (sum of wgt is 8.2331e+09) Linear regression Number of obs = 336 F(4, 331) = 30.14 Prob > F = 0.0000 R-squared = 0.2833 Root MSE = .30825 -------------------------------------------------------------------------------- | Robust lntwtrateabort | Coef. Std. Err. t P>|t| [95% Conf. Interval] ---------------+---------------------------------------------------------------- broadcastday | .1589494 .0390089 4.07 0.000 .0822128 .2356859 L_broadcastday | .2156727 .0492342 4.38 0.000 .1188213 .3125241 trend | -.0002574 .0002711 -0.95 0.343 -.0007906 .0002759 trend2 | -1.92e-07 1.61e-07 -1.19 0.233 -5.09e-07 1.25e-07 _cons | 4.781726 .1127923 42.39 0.000 4.559846 5.003606 -------------------------------------------------------------------------------- . . **** . * Column 8: Abortion tweets, Replicate KL Table 2, column 5 with corrected dates and lag varia > bles, "in season" only . reg lntwtrateabort r_broadcastday L_r_broadcastday trend trend2 [weight=totaltweets] if r_insea > son==1, robust (analytic weights assumed) (sum of wgt is 8.1501e+09) Linear regression Number of obs = 333 F(4, 328) = 22.80 Prob > F = 0.0000 R-squared = 0.2498 Root MSE = .31653 ---------------------------------------------------------------------------------- | Robust lntwtrateabort | Coef. Std. Err. t P>|t| [95% Conf. Interval] -----------------+---------------------------------------------------------------- r_broadcastday | .0897429 .0400845 2.24 0.026 .0108878 .1685981 L_r_broadcastday | .1403409 .0470953 2.98 0.003 .0476939 .2329879 trend | -.0001987 .0002777 -0.72 0.475 -.000745 .0003475 trend2 | -2.30e-07 1.67e-07 -1.38 0.170 -5.58e-07 9.89e-08 _cons | 4.785016 .1147595 41.70 0.000 4.559258 5.010773 ---------------------------------------------------------------------------------- . . **** . * Column 9: Abortion tweets, Replicate KL Table 2, column 5 with corrected dates and lag varia > bles, "in season" only, no weights . reg lntwtrateabort r_broadcastday L_r_broadcastday trend trend2 if r_inseason==1, robust Linear regression Number of obs = 333 F(4, 328) = 26.43 Prob > F = 0.0000 R-squared = 0.1762 Root MSE = .40879 ---------------------------------------------------------------------------------- | Robust lntwtrateabort | Coef. Std. Err. t P>|t| [95% Conf. Interval] -----------------+---------------------------------------------------------------- r_broadcastday | .1678003 .0484033 3.47 0.001 .0725802 .2630203 L_r_broadcastday | .1703923 .0582092 2.93 0.004 .0558817 .2849028 trend | .0002549 .0002643 0.96 0.336 -.0002651 .0007748 trend2 | -4.96e-07 1.72e-07 -2.89 0.004 -8.33e-07 -1.58e-07 _cons | 4.58358 .0972113 47.15 0.000 4.392344 4.774817 ---------------------------------------------------------------------------------- . . . **** . * Column 10: Abortion tweets, Corrected dates and correct lag variables, all dates . reg lntwtrateabort r_broadcastday L_r_broadcastday trend trend2 pre16p r_outofseason [weight=to > taltweets], robust (analytic weights assumed) (sum of wgt is 3.6712e+10) Linear regression Number of obs = 1,455 F(6, 1448) = 37.03 Prob > F = 0.0000 R-squared = 0.0614 Root MSE = .44809 ---------------------------------------------------------------------------------- | Robust lntwtrateabort | Coef. Std. Err. t P>|t| [95% Conf. Interval] -----------------+---------------------------------------------------------------- r_broadcastday | .0946055 .0477446 1.98 0.048 .0009496 .1882614 L_r_broadcastday | .1580955 .048514 3.26 0.001 .0629303 .2532607 trend | -.0006779 .0002234 -3.03 0.002 -.001116 -.0002397 trend2 | 2.24e-07 1.36e-07 1.64 0.100 -4.31e-08 4.91e-07 pre16p | .3262872 .0884213 3.69 0.000 .1528396 .4997347 r_outofseason | .1967221 .030124 6.53 0.000 .1376308 .2558134 _cons | 4.773396 .0852672 55.98 0.000 4.606136 4.940657 ---------------------------------------------------------------------------------- . lincom r_broadcastday - pre16p ( 1) r_broadcastday - pre16p = 0 ------------------------------------------------------------------------------ lntwtratea~t | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- (1) | -.2316817 .0940512 -2.46 0.014 -.4161729 -.0471905 ------------------------------------------------------------------------------ . lincom L_r_broadcastday - pre16p ( 1) L_r_broadcastday - pre16p = 0 ------------------------------------------------------------------------------ lntwtratea~t | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- (1) | -.1681917 .0945672 -1.78 0.076 -.3536951 .0173117 ------------------------------------------------------------------------------ . lincom r_broadcastday - r_outofseason ( 1) r_broadcastday - r_outofseason = 0 ------------------------------------------------------------------------------ lntwtratea~t | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- (1) | -.1021166 .0442262 -2.31 0.021 -.1888708 -.0153624 ------------------------------------------------------------------------------ . lincom L_r_broadcastday - r_outofseason ( 1) L_r_broadcastday - r_outofseason = 0 ------------------------------------------------------------------------------ lntwtratea~t | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- (1) | -.0386266 .0449626 -0.86 0.390 -.1268253 .0495722 ------------------------------------------------------------------------------ . . **** . * Column 11: Abortion tweets, Corrected dates and correct lag variables, all dates . newey lntwtrateabort r_broadcastday L_r_broadcastday trend trend2 pre16p r_outofseason, lag(1) > force Regression with Newey-West standard errors Number of obs = 1,455 maximum lag: 1 F( 6, 1448) = 17.10 Prob > F = 0.0000 ---------------------------------------------------------------------------------- | Newey-West lntwtrateabort | Coef. Std. Err. t P>|t| [95% Conf. Interval] -----------------+---------------------------------------------------------------- r_broadcastday | .1717631 .0529331 3.24 0.001 .0679293 .2755969 L_r_broadcastday | .1839348 .0574056 3.20 0.001 .0713277 .2965419 trend | -.0001076 .0002218 -0.49 0.628 -.0005427 .0003275 trend2 | -8.92e-08 1.36e-07 -0.66 0.511 -3.55e-07 1.77e-07 pre16p | .4175797 .1000052 4.18 0.000 .2214091 .6137503 r_outofseason | .152572 .0412272 3.70 0.000 .0717006 .2334434 _cons | 4.581894 .0908796 50.42 0.000 4.403624 4.760164 ---------------------------------------------------------------------------------- . lincom r_broadcastday - pre16p ( 1) r_broadcastday - pre16p = 0 ------------------------------------------------------------------------------ lntwtratea~t | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- (1) | -.2458166 .100808 -2.44 0.015 -.443562 -.0480712 ------------------------------------------------------------------------------ . lincom L_r_broadcastday - pre16p ( 1) L_r_broadcastday - pre16p = 0 ------------------------------------------------------------------------------ lntwtratea~t | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- (1) | -.2336449 .1039933 -2.25 0.025 -.4376386 -.0296512 ------------------------------------------------------------------------------ . lincom r_broadcastday - r_outofseason ( 1) r_broadcastday - r_outofseason = 0 ------------------------------------------------------------------------------ lntwtratea~t | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- (1) | .0191911 .0472986 0.41 0.685 -.0735901 .1119723 ------------------------------------------------------------------------------ . lincom L_r_broadcastday - r_outofseason ( 1) L_r_broadcastday - r_outofseason = 0 ------------------------------------------------------------------------------ lntwtratea~t | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- (1) | .0313628 .0521337 0.60 0.548 -.0709029 .1336285 ------------------------------------------------------------------------------ . . ************ . . . . . ************ . * Table 2 Results: Association between Log(16P Tweets) and Twitter Activity . . . . **** . * Column 1 Birth control tweets, Replicate KL Table 4, column 4 . reg lntwtratebc lntwtrate16p trend trend2 [weight=totaltweets] if r_inseason==1, robust (analytic weights assumed) (sum of wgt is 8.1501e+09) Linear regression Number of obs = 333 F(3, 329) = 7.17 Prob > F = 0.0001 R-squared = 0.0458 Root MSE = .32213 ------------------------------------------------------------------------------ | Robust lntwtratebc | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- lntwtrate16p | .0774935 .0344215 2.25 0.025 .0097795 .1452076 trend | .0007188 .000269 2.67 0.008 .0001896 .0012479 trend2 | -3.45e-07 1.74e-07 -1.98 0.048 -6.86e-07 -2.96e-09 _cons | 3.105683 .1426058 21.78 0.000 2.825148 3.386217 ------------------------------------------------------------------------------ . . **** . * Column 2 Birth control tweets, Replicate KL Table 4, column 4 without weights . reg lntwtratebc lntwtrate16p trend trend2 if r_inseason==1, robust Linear regression Number of obs = 333 F(3, 329) = 26.79 Prob > F = 0.0000 R-squared = 0.1408 Root MSE = .35017 ------------------------------------------------------------------------------ | Robust lntwtratebc | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- lntwtrate16p | .0284664 .0257522 1.11 0.270 -.0221933 .0791261 trend | .001566 .0002305 6.79 0.000 .0011126 .0020194 trend2 | -8.93e-07 1.59e-07 -5.61 0.000 -1.21e-06 -5.80e-07 _cons | 2.99426 .0962869 31.10 0.000 2.804845 3.183676 ------------------------------------------------------------------------------ . . . **** . * Column 3 Birth control tweets, all dates . reg lntwtratebc lntwtrate16p trend trend2 [weight=totaltweets] if pre16p==0, robust (analytic weights assumed) (sum of wgt is 3.6453e+10) Linear regression Number of obs = 1,292 F(3, 1288) = 35.28 Prob > F = 0.0000 R-squared = 0.0377 Root MSE = .496 ------------------------------------------------------------------------------ | Robust lntwtratebc | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- lntwtrate16p | .0344417 .0178424 1.93 0.054 -.0005617 .0694451 trend | .0007691 .0002469 3.11 0.002 .0002846 .0012535 trend2 | -2.80e-07 1.44e-07 -1.95 0.051 -5.62e-07 1.37e-09 _cons | 3.278298 .0828928 39.55 0.000 3.115679 3.440918 ------------------------------------------------------------------------------ . . . **** . * Column 4 Birth control tweets, all dates, Newey-West . newey lntwtratebc lntwtrate16p trend trend2 if pre16p==0, lag(1) force Regression with Newey-West standard errors Number of obs = 1,292 maximum lag: 1 F( 3, 1288) = 27.77 Prob > F = 0.0000 ------------------------------------------------------------------------------ | Newey-West lntwtratebc | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- lntwtrate16p | .0101449 .011222 0.90 0.366 -.0118705 .0321603 trend | .0008062 .0001894 4.26 0.000 .0004347 .0011777 trend2 | -3.00e-07 1.15e-07 -2.61 0.009 -5.25e-07 -7.45e-08 _cons | 3.313075 .0629303 52.65 0.000 3.189618 3.436533 ------------------------------------------------------------------------------ . . . **** . * Column 5 Abortion tweets, Replicate KL Table 4, column 4 . reg lntwtrateabort lntwtrate16p trend trend2 [weight=totaltweets] if r_inseason==1, robust (analytic weights assumed) (sum of wgt is 8.1501e+09) Linear regression Number of obs = 333 F(3, 329) = 30.83 Prob > F = 0.0000 R-squared = 0.2391 Root MSE = .3183 ------------------------------------------------------------------------------ | Robust lntwtratea~t | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- lntwtrate16p | .0599061 .0247682 2.42 0.016 .011182 .1086302 trend | -.0002008 .0002818 -0.71 0.477 -.0007552 .0003535 trend2 | -2.25e-07 1.69e-07 -1.33 0.184 -5.57e-07 1.07e-07 _cons | 4.615126 .1426541 32.35 0.000 4.334497 4.895755 ------------------------------------------------------------------------------ . . **** . * Column 6 Abortion tweets, Replicate KL Table 4, column 4, without weights . reg lntwtrateabort lntwtrate16p trend trend2 if r_inseason==1, robust Linear regression Number of obs = 333 F(3, 329) = 24.22 Prob > F = 0.0000 R-squared = 0.1464 Root MSE = .41548 ------------------------------------------------------------------------------ | Robust lntwtratea~t | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- lntwtrate16p | .0052017 .0295733 0.18 0.860 -.052975 .0633784 trend | .000267 .0002717 0.98 0.326 -.0002675 .0008016 trend2 | -5.06e-07 1.76e-07 -2.87 0.004 -8.53e-07 -1.59e-07 _cons | 4.613844 .1377774 33.49 0.000 4.342808 4.884879 ------------------------------------------------------------------------------ . . **** . * Column 7 Abortion tweets, all dates . reg lntwtrateabort lntwtrate16p trend trend2 [weight=totaltweets] if pre16p==0, robust (analytic weights assumed) (sum of wgt is 3.6453e+10) Linear regression Number of obs = 1,292 F(3, 1288) = 25.87 Prob > F = 0.0000 R-squared = 0.0319 Root MSE = .45127 ------------------------------------------------------------------------------ | Robust lntwtratea~t | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- lntwtrate16p | -.0218331 .0177588 -1.23 0.219 -.0566724 .0130061 trend | -.0006961 .0002226 -3.13 0.002 -.0011328 -.0002593 trend2 | 2.58e-07 1.32e-07 1.95 0.051 -1.15e-09 5.18e-07 _cons | 4.971381 .0796618 62.41 0.000 4.8151 5.127662 ------------------------------------------------------------------------------ . . . **** . * Column 8 Abortion tweets, all dates, Newey-West . newey lntwtrateabort lntwtrate16p trend trend2 if pre16p==0, lag(1) force Regression with Newey-West standard errors Number of obs = 1,292 maximum lag: 1 F( 3, 1288) = 9.92 Prob > F = 0.0000 ------------------------------------------------------------------------------ | Newey-West lntwtratea~t | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- lntwtrate16p | -.0186488 .0155652 -1.20 0.231 -.0491847 .011887 trend | -.0001349 .000227 -0.59 0.552 -.0005801 .0003104 trend2 | -5.44e-08 1.36e-07 -0.40 0.690 -3.22e-07 2.13e-07 _cons | 4.750216 .082686 57.45 0.000 4.588002 4.91243 ------------------------------------------------------------------------------ . . . clear . . . ************ . . ************ . * Data: Table 3, State-level Twitter trends . * Data file obtained from https://www.aeaweb.org/aer/data/10512/20140012_data.zip . use "$work/data/topsy-state-daily.dta" . . **** . * Calculate total tweets by state and period . sort stname period . collapse (sum) totaltweets, by(stname period) . sort stname period . save "$work/temp.dta", replace (note: file /Users/djaeger/Dropbox/research/SixteenAndPregnant/6docs/manuscript/IREE/replication/ > /temp.dta not found) file ~/Dropbox/research/SixteenAndPregnant/6docs/manuscript/IREE/replication//temp.dta saved . . clear . . **** . . **** . * Merge total tweets into state-by-period dataset . use "$work/data/topsy-state.dta" . drop totaltweets . drop if stname == "PR" (11 observations deleted) . sort stname period . merge 1:1 stname period using "$work/temp.dta" Result # of obs. ----------------------------------------- not matched 0 matched 561 (_merge==3) ----------------------------------------- . rm "$work/temp.dta" . . **** . . **** . * Generate tweet rate variables . . gen twtrate16p = 1000000*tweets16p/totaltweets . gen twtratebc = 1000000*tweetsbc/totaltweets . gen twtrateabort = 1000000*tweetsabort/totaltweets . gen twtrateadopt = 1000000*tweetsadopt/totaltweets . gen lntwtrate16p = log(twtrate16p) (19 missing values generated) . gen lntwtratebc = log(twtratebc) (6 missing values generated) . gen lntwtrateabort = log(twtrateabort) (1 missing value generated) . gen lntwtrateadopt = log(twtrateadopt) (5 missing values generated) . . **** . . **** . * drop observations if missing any of relevant tweet variables . . drop if twtrate16p == 0 | twtrateabort == 0 | twtratebc == 0 (24 observations deleted) . . **** . . . . **************** . * Table 3, State-level Twitter Results . . gen logtotaltweets=log(totaltweets) . . **** . * Column 1: Association between tweet rate for birth control and tweet rate for 16 and pregnan > t, weighted . areg lntwtratebc lntwtrate16p i.period [aw=totaltweets], absorb(stname) robust cluster(stname) (sum of wgt is 1.2847e+10) Linear regression, absorbing indicators Number of obs = 537 F( 11, 50) = 127.49 Prob > F = 0.0000 R-squared = 0.8448 Adj R-squared = 0.8249 Root MSE = 0.1580 (Std. Err. adjusted for 51 clusters in stname) ------------------------------------------------------------------------------ | Robust lntwtratebc | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- lntwtrate16p | .1367762 .0542839 2.52 0.015 .0277438 .2458085 | period | 2 | -.6378096 .2178753 -2.93 0.005 -1.075425 -.2001942 3 | .0369613 .0843711 0.44 0.663 -.132503 .2064256 4 | -.2427254 .2098769 -1.16 0.253 -.6642755 .1788248 5 | .2673982 .1454028 1.84 0.072 -.0246519 .5594482 6 | -.2666657 .1993489 -1.34 0.187 -.6670698 .1337384 7 | -.0612432 .1371924 -0.45 0.657 -.3368023 .214316 8 | -.2329832 .18533 -1.26 0.215 -.6052295 .1392631 9 | .72051 .1704127 4.23 0.000 .3782261 1.062794 10 | .1783689 .1960578 0.91 0.367 -.2154248 .5721625 11 | .4449588 .1476663 3.01 0.004 .1483624 .7415553 | _cons | 3.521577 .0272058 129.44 0.000 3.466932 3.576221 -------------+---------------------------------------------------------------- stname | absorbed (51 categories) . . . **** . * Column 2: Association between tweet rate for birth control and tweet rate for 16 and pregnan > t, unweighted . areg lntwtratebc lntwtrate16p i.period, absorb(stname) robust cluster(stname) Linear regression, absorbing indicators Number of obs = 537 F( 11, 50) = 96.80 Prob > F = 0.0000 R-squared = 0.6737 Adj R-squared = 0.6318 Root MSE = 0.3077 (Std. Err. adjusted for 51 clusters in stname) ------------------------------------------------------------------------------ | Robust lntwtratebc | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- lntwtrate16p | .0485518 .0476308 1.02 0.313 -.0471175 .1442212 | period | 2 | -.3123715 .209898 -1.49 0.143 -.733964 .1092211 3 | .12769 .0812685 1.57 0.122 -.0355425 .2909225 4 | .1808734 .2150191 0.84 0.404 -.2510052 .612752 5 | .581037 .1568572 3.70 0.001 .2659801 .896094 6 | .0591899 .1711969 0.35 0.731 -.2846693 .403049 7 | .0716334 .1232429 0.58 0.564 -.1759071 .319174 8 | .0550938 .1601369 0.34 0.732 -.2665506 .3767382 9 | 1.022861 .1589643 6.43 0.000 .7035717 1.34215 10 | .5601142 .1901115 2.95 0.005 .178264 .9419645 11 | .8304286 .1468574 5.65 0.000 .5354569 1.1254 | _cons | 3.537374 .0308217 114.77 0.000 3.475467 3.599281 -------------+---------------------------------------------------------------- stname | absorbed (51 categories) . . . **** . * Column 3: Association between tweet rate for abortion and tweet rate for 16 and pregnant, we > ighted . areg lntwtrateabort lntwtrate16p i.period [aw=totaltweets], absorb(stname) robust cluster(stnam > e) (sum of wgt is 1.2847e+10) Linear regression, absorbing indicators Number of obs = 537 F( 11, 50) = 142.54 Prob > F = 0.0000 R-squared = 0.7499 Adj R-squared = 0.7178 Root MSE = 0.1824 (Std. Err. adjusted for 51 clusters in stname) ------------------------------------------------------------------------------ | Robust lntwtrate~rt | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- lntwtrate16p | -.0870339 .0754457 -1.15 0.254 -.238571 .0645031 | period | 2 | .0857033 .3073064 0.28 0.781 -.5315396 .7029463 3 | .1525114 .1073114 1.42 0.161 -.0630298 .3680526 4 | .5686094 .2860504 1.99 0.052 -.0059396 1.143159 5 | -.0079726 .2156494 -0.04 0.971 -.4411171 .4251719 6 | -.3262031 .2859583 -1.14 0.259 -.9005672 .248161 7 | .3737025 .2082346 1.79 0.079 -.0445489 .791954 8 | -.1219591 .2787233 -0.44 0.664 -.6817913 .437873 9 | -.0450582 .2332398 -0.19 0.848 -.5135341 .4234178 10 | -.2023884 .2725546 -0.74 0.461 -.7498303 .3450536 11 | .1506058 .1874825 0.80 0.426 -.2259638 .5271754 | _cons | 5.133423 .0407487 125.98 0.000 5.051577 5.215269 -------------+---------------------------------------------------------------- stname | absorbed (51 categories) . . . **** . * Column 4: Association between tweet rate for abortion and tweet rate for 16 and pregnant, un > weighted . areg lntwtrateabort lntwtrate16p i.period, absorb(stname) robust cluster(stname) Linear regression, absorbing indicators Number of obs = 537 F( 11, 50) = 54.15 Prob > F = 0.0000 R-squared = 0.6950 Adj R-squared = 0.6559 Root MSE = 0.3208 (Std. Err. adjusted for 51 clusters in stname) ------------------------------------------------------------------------------ | Robust lntwtrate~rt | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- lntwtrate16p | .0106319 .0413468 0.26 0.798 -.0724155 .0936793 | period | 2 | -.2974772 .1641976 -1.81 0.076 -.6272778 .0323235 3 | .0450796 .0846792 0.53 0.597 -.1250035 .2151628 4 | .2395791 .1807749 1.33 0.191 -.1235179 .6026761 5 | -.2241355 .1397242 -1.60 0.115 -.5047799 .0565089 6 | -.8728389 .1726862 -5.05 0.000 -1.219689 -.5259885 7 | .0264636 .1297 0.20 0.839 -.2340466 .2869738 8 | -.6398401 .16279 -3.93 0.000 -.9668134 -.3128668 9 | -.4673348 .1407116 -3.32 0.002 -.7499623 -.1847072 10 | -.7178464 .1695929 -4.23 0.000 -1.058484 -.377209 11 | -.0738429 .1404407 -0.53 0.601 -.3559264 .2082405 | _cons | 5.25289 .0456218 115.14 0.000 5.161256 5.344524 -------------+---------------------------------------------------------------- stname | absorbed (51 categories) . . clear . . **************** . . . . ************ . * Data: Table 4, Google Results . * Data file obtained directly from Phil Levine on 28 August 2016 . * This file is referenced in KL's replication files on the AER website, but not . * included in the downloadable zip file. . * . * Variables: . * new16p is an indicator variable for week with new 16 and Pregnant broadcast. . * index16p is the Google Trends index for 16 and Pregnant . * howgetbc is the Google Trends index for "How to get birth control" . * howgetabort is the Google Trends index for "How to get an abortion" . * howgetpill is the Google Trends index for "How to get birth control pill" . . use "$work/data/googletrends.dta" . tsset trend time variable: trend, 1 to 209 delta: 1 unit . . . ************ . . . . ************ . * Table 4: Results . . **** . * Column 1: Association between Google searches for 16P and 16P broadcast week . reg index16p new16p trend trend2 i.month Source | SS df MS Number of obs = 209 -------------+---------------------------------- F(14, 194) = 63.87 Model | 101394.489 14 7242.46352 Prob > F = 0.0000 Residual | 21998.6016 194 113.394854 R-squared = 0.8217 -------------+---------------------------------- Adj R-squared = 0.8089 Total | 123393.091 208 593.236014 Root MSE = 10.649 ------------------------------------------------------------------------------ index16p | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- new16p | 38.95436 1.975252 19.72 0.000 35.05864 42.85009 trend | .7525006 .0505813 14.88 0.000 .6527407 .8522605 trend2 | -.003203 .0002335 -13.72 0.000 -.0036635 -.0027425 | month | 2 | -4.329551 3.623921 -1.19 0.234 -11.47689 2.81779 3 | -7.771047 3.589884 -2.16 0.032 -14.85126 -.690836 4 | -3.880917 3.711969 -1.05 0.297 -11.20191 3.440079 5 | -14.75121 3.552389 -4.15 0.000 -21.75747 -7.744946 6 | -15.04519 3.754076 -4.01 0.000 -22.44923 -7.641143 7 | -6.154802 3.534865 -1.74 0.083 -13.1265 .8168973 8 | -11.61247 3.514575 -3.30 0.001 -18.54415 -4.680788 9 | -11.65643 3.584306 -3.25 0.001 -18.72564 -4.58722 10 | -13.93203 3.532323 -3.94 0.000 -20.89872 -6.965346 11 | -11.47685 3.600854 -3.19 0.002 -18.5787 -4.375003 12 | -7.874053 3.632276 -2.17 0.031 -15.03787 -.7102324 | _cons | .2957238 3.163176 0.09 0.926 -5.942906 6.534354 ------------------------------------------------------------------------------ . newey index16p new16p trend trend2 i.month, lag(1) force Regression with Newey-West standard errors Number of obs = 209 maximum lag: 1 F( 14, 194) = 69.93 Prob > F = 0.0000 ------------------------------------------------------------------------------ | Newey-West index16p | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- new16p | 38.95436 2.458314 15.85 0.000 34.10591 43.80282 trend | .7525006 .0477181 15.77 0.000 .6583877 .8466135 trend2 | -.003203 .0002192 -14.61 0.000 -.0036352 -.0027707 | month | 2 | -4.329551 5.333398 -0.81 0.418 -14.84844 6.189337 3 | -7.771047 4.60282 -1.69 0.093 -16.84904 1.306945 4 | -3.880917 5.831405 -0.67 0.507 -15.38201 7.620174 5 | -14.75121 4.6826 -3.15 0.002 -23.98655 -5.515868 6 | -15.04519 5.370138 -2.80 0.006 -25.63653 -4.453837 7 | -6.154802 4.976489 -1.24 0.218 -15.96977 3.660165 8 | -11.61247 4.649572 -2.50 0.013 -20.78267 -2.442269 9 | -11.65643 4.826917 -2.41 0.017 -21.1764 -2.136458 10 | -13.93203 4.878854 -2.86 0.005 -23.55444 -4.309625 11 | -11.47685 5.792157 -1.98 0.049 -22.90053 -.0531661 12 | -7.874053 4.748758 -1.66 0.099 -17.23987 1.491769 | _cons | .2957238 4.201241 0.07 0.944 -7.990248 8.581695 ------------------------------------------------------------------------------ . . **** . * Column 2: Association between Google searches for birth control and 16P broadcast week . reg howgetbc new16p trend trend2 i.month Source | SS df MS Number of obs = 209 -------------+---------------------------------- F(14, 194) = 46.35 Model | 27886.4377 14 1991.88841 Prob > F = 0.0000 Residual | 8337.48577 194 42.9767308 R-squared = 0.7698 -------------+---------------------------------- Adj R-squared = 0.7532 Total | 36223.9234 208 174.153478 Root MSE = 6.5557 ------------------------------------------------------------------------------ howgetbc | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- new16p | .8254411 1.216025 0.68 0.498 -1.572886 3.223768 trend | .3932417 .0311394 12.63 0.000 .3318264 .4546569 trend2 | -.0010298 .0001437 -7.16 0.000 -.0013133 -.0007463 | month | 2 | -4.993157 2.230996 -2.24 0.026 -9.393278 -.5930371 3 | -5.806896 2.210041 -2.63 0.009 -10.16569 -1.448103 4 | -5.102503 2.285201 -2.23 0.027 -9.609531 -.5954759 5 | -5.175152 2.186958 -2.37 0.019 -9.488419 -.8618844 6 | .8336057 2.311123 0.36 0.719 -3.724548 5.391759 7 | 3.560199 2.17617 1.64 0.103 -.7317905 7.852189 8 | -2.500788 2.163679 -1.16 0.249 -6.768142 1.766566 9 | -5.083575 2.206607 -2.30 0.022 -9.435595 -.7315545 10 | -8.147372 2.174605 -3.75 0.000 -12.43628 -3.858469 11 | -7.326691 2.216795 -3.31 0.001 -11.6988 -2.954578 12 | -7.937928 2.236139 -3.55 0.000 -12.34819 -3.527663 | _cons | 47.57063 1.947347 24.43 0.000 43.72994 51.41132 ------------------------------------------------------------------------------ . newey howgetbc new16p trend trend2 i.month, lag(1) force Regression with Newey-West standard errors Number of obs = 209 maximum lag: 1 F( 14, 194) = 59.77 Prob > F = 0.0000 ------------------------------------------------------------------------------ | Newey-West howgetbc | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- new16p | .8254411 1.496777 0.55 0.582 -2.126603 3.777485 trend | .3932417 .031808 12.36 0.000 .3305077 .4559756 trend2 | -.0010298 .0001545 -6.66 0.000 -.0013346 -.000725 | month | 2 | -4.993157 2.300211 -2.17 0.031 -9.52979 -.4565251 3 | -5.806896 2.276723 -2.55 0.012 -10.2972 -1.316589 4 | -5.102503 2.364903 -2.16 0.032 -9.766724 -.4382824 5 | -5.175152 2.368079 -2.19 0.030 -9.845637 -.5046666 6 | .8336057 2.752932 0.30 0.762 -4.595912 6.263123 7 | 3.560199 2.899697 1.23 0.221 -2.158778 9.279177 8 | -2.500788 2.211721 -1.13 0.260 -6.862894 1.861318 9 | -5.083575 2.729515 -1.86 0.064 -10.46691 .2997594 10 | -8.147372 2.23607 -3.64 0.000 -12.5575 -3.737243 11 | -7.326691 2.580275 -2.84 0.005 -12.41568 -2.237697 12 | -7.937928 2.840384 -2.79 0.006 -13.53993 -2.335932 | _cons | 47.57063 2.079889 22.87 0.000 43.46853 51.67272 ------------------------------------------------------------------------------ . . **** . * Column 3: Association between Google searches for birth control and Google searches for 16P . * Replicates KL Table 4, Panel A, column 1 . reg howgetbc index16p trend trend2 i.month Source | SS df MS Number of obs = 209 -------------+---------------------------------- F(14, 194) = 46.27 Model | 27875.6408 14 1991.1172 Prob > F = 0.0000 Residual | 8348.28265 194 43.0323848 R-squared = 0.7695 -------------+---------------------------------- Adj R-squared = 0.7529 Total | 36223.9234 208 174.153478 Root MSE = 6.5599 ------------------------------------------------------------------------------ howgetbc | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- index16p | .0116722 .0255149 0.46 0.648 -.03865 .0619945 trend | .3865975 .0392857 9.84 0.000 .3091156 .4640793 trend2 | -.0010025 .0001773 -5.65 0.000 -.0013523 -.0006528 | month | 2 | -4.890889 2.226153 -2.20 0.029 -9.281458 -.5003201 3 | -5.623502 2.191025 -2.57 0.011 -9.94479 -1.302215 4 | -4.859254 2.234536 -2.17 0.031 -9.266356 -.4521509 5 | -4.849757 2.130096 -2.28 0.024 -9.050876 -.6486389 6 | 1.195573 2.233276 0.54 0.593 -3.209045 5.60019 7 | 3.692611 2.168492 1.70 0.090 -.584236 7.969458 8 | -2.387593 2.192959 -1.09 0.278 -6.712696 1.937509 9 | -4.965914 2.234278 -2.22 0.027 -9.372508 -.5593207 10 | -7.966343 2.196608 -3.63 0.000 -12.29864 -3.634043 11 | -7.12403 2.209414 -3.22 0.001 -11.48159 -2.766473 12 | -7.771528 2.224191 -3.49 0.001 -12.15823 -3.384829 | _cons | 47.51113 1.945148 24.43 0.000 43.67478 51.34748 ------------------------------------------------------------------------------ . newey howgetbc index16p trend trend2 i.month, lag(1) force Regression with Newey-West standard errors Number of obs = 209 maximum lag: 1 F( 14, 194) = 59.94 Prob > F = 0.0000 ------------------------------------------------------------------------------ | Newey-West howgetbc | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- index16p | .0116722 .0304139 0.38 0.702 -.0483121 .0716565 trend | .3865975 .0410031 9.43 0.000 .3057283 .4674666 trend2 | -.0010025 .0001915 -5.23 0.000 -.0013803 -.0006248 | month | 2 | -4.890889 2.294308 -2.13 0.034 -9.415877 -.3659002 3 | -5.623502 2.25611 -2.49 0.014 -10.07316 -1.173849 4 | -4.859254 2.28741 -2.12 0.035 -9.370639 -.3478684 5 | -4.849757 2.271591 -2.13 0.034 -9.329943 -.3695721 6 | 1.195573 2.64585 0.45 0.652 -4.022751 6.413897 7 | 3.692611 2.849914 1.30 0.197 -1.928182 9.313404 8 | -2.387593 2.213235 -1.08 0.282 -6.752684 1.977497 9 | -4.965914 2.709911 -1.83 0.068 -10.31058 .378755 10 | -7.966343 2.244961 -3.55 0.000 -12.39401 -3.538678 11 | -7.12403 2.556563 -2.79 0.006 -12.16626 -2.081803 12 | -7.771528 2.876905 -2.70 0.008 -13.44556 -2.097501 | _cons | 47.51113 2.099832 22.63 0.000 43.3697 51.65256 ------------------------------------------------------------------------------ . . **** . * Column 4: Association between Google searches for birth control pill and 16P broadcast week . reg howgetpill new16p trend trend2 i.month Source | SS df MS Number of obs = 209 -------------+---------------------------------- F(14, 194) = 4.03 Model | 6616.62965 14 472.616404 Prob > F = 0.0000 Residual | 22771.0928 194 117.376767 R-squared = 0.2251 -------------+---------------------------------- Adj R-squared = 0.1692 Total | 29387.7225 208 141.287127 Root MSE = 10.834 ------------------------------------------------------------------------------ howgetpill | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- new16p | 2.226714 2.009634 1.11 0.269 -1.736821 6.19025 trend | .157101 .0514618 3.05 0.003 .0556047 .2585974 trend2 | -.0004282 .0002376 -1.80 0.073 -.0008967 .0000403 | month | 2 | -.0368611 3.687 -0.01 0.992 -7.308611 7.234889 3 | -1.141239 3.65237 -0.31 0.755 -8.34469 6.062212 4 | 7.790248 3.776581 2.06 0.040 .3418204 15.23868 5 | 2.654162 3.614223 0.73 0.464 -4.474052 9.782377 6 | 4.663185 3.819421 1.22 0.224 -2.869734 12.1961 7 | 6.465448 3.596394 1.80 0.074 -.6276023 13.5585 8 | 2.24695 3.57575 0.63 0.530 -4.805387 9.299286 9 | -2.356839 3.646695 -0.65 0.519 -9.549097 4.835419 10 | -3.659154 3.593807 -1.02 0.310 -10.7471 3.428796 11 | -.3257929 3.663531 -0.09 0.929 -7.551256 6.89967 12 | -.512084 3.6955 -0.14 0.890 -7.8006 6.776432 | _cons | 43.1157 3.218235 13.40 0.000 36.76848 49.46292 ------------------------------------------------------------------------------ . newey howgetpill new16p trend trend2 i.month, lag(1) force Regression with Newey-West standard errors Number of obs = 209 maximum lag: 1 F( 14, 194) = 4.66 Prob > F = 0.0000 ------------------------------------------------------------------------------ | Newey-West howgetpill | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- new16p | 2.226714 2.385493 0.93 0.352 -2.478117 6.931545 trend | .157101 .04444 3.54 0.001 .0694534 .2447486 trend2 | -.0004282 .000217 -1.97 0.050 -.0008562 -1.98e-07 | month | 2 | -.0368611 3.242576 -0.01 0.991 -6.432088 6.358366 3 | -1.141239 4.165943 -0.27 0.784 -9.357593 7.075115 4 | 7.790248 4.993872 1.56 0.120 -2.059003 17.6395 5 | 2.654162 3.30817 0.80 0.423 -3.870434 9.178758 6 | 4.663185 5.355462 0.87 0.385 -5.899218 15.22559 7 | 6.465448 3.46333 1.87 0.063 -.3651644 13.29606 8 | 2.24695 3.665927 0.61 0.541 -4.983238 9.477138 9 | -2.356839 2.953343 -0.80 0.426 -8.181622 3.467944 10 | -3.659154 3.802686 -0.96 0.337 -11.15907 3.84076 11 | -.3257929 3.186961 -0.10 0.919 -6.611332 5.959746 12 | -.512084 3.371687 -0.15 0.879 -7.161952 6.137784 | _cons | 43.1157 2.768528 15.57 0.000 37.65542 48.57597 ------------------------------------------------------------------------------ . . **** . * Column 5: Association between Google searches for birth control pill and Google searches for > 16P . * Replicates KL Table 4, Panel A, column 2 . reg howgetpill index16p trend trend2 i.month Source | SS df MS Number of obs = 209 -------------+---------------------------------- F(14, 194) = 4.16 Model | 6783.04184 14 484.502989 Prob > F = 0.0000 Residual | 22604.6806 194 116.518972 R-squared = 0.2308 -------------+---------------------------------- Adj R-squared = 0.1753 Total | 29387.7225 208 141.287127 Root MSE = 10.794 ------------------------------------------------------------------------------ howgetpill | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- index16p | .0685392 .0419851 1.63 0.104 -.0142666 .151345 trend | .1029682 .0646449 1.59 0.113 -.0245289 .2304653 trend2 | -.0001965 .0002918 -0.67 0.501 -.0007721 .000379 | month | 2 | .1980434 3.663157 0.05 0.957 -7.026683 7.42277 3 | -.7194132 3.605354 -0.20 0.842 -7.830136 6.39131 4 | 7.819621 3.676952 2.13 0.035 .5676881 15.07155 5 | 3.482052 3.505094 0.99 0.322 -3.430932 10.39504 6 | 5.471608 3.674879 1.49 0.138 -1.776236 12.71945 7 | 6.814889 3.568277 1.91 0.058 -.2227068 13.85249 8 | 3.069574 3.608537 0.85 0.396 -4.047426 10.18657 9 | -1.535927 3.676527 -0.42 0.677 -8.787022 5.715168 10 | -2.726272 3.614542 -0.75 0.452 -9.855115 4.402572 11 | .3786996 3.635614 0.10 0.917 -6.791704 7.549103 12 | -.0614476 3.659929 -0.02 0.987 -7.279807 7.156912 | _cons | 43.16242 3.200761 13.49 0.000 36.84967 49.47518 ------------------------------------------------------------------------------ . newey howgetpill index16p trend trend2 i.month, lag(1) force Regression with Newey-West standard errors Number of obs = 209 maximum lag: 1 F( 14, 194) = 4.94 Prob > F = 0.0000 ------------------------------------------------------------------------------ | Newey-West howgetpill | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- index16p | .0685392 .0479064 1.43 0.154 -.0259451 .1630235 trend | .1029682 .0616409 1.67 0.096 -.0186041 .2245404 trend2 | -.0001965 .0002831 -0.69 0.488 -.0007548 .0003617 | month | 2 | .1980434 3.268132 0.06 0.952 -6.247588 6.643674 3 | -.7194132 4.33003 -0.17 0.868 -9.259392 7.820565 4 | 7.819621 4.950866 1.58 0.116 -1.944811 17.58405 5 | 3.482052 3.227753 1.08 0.282 -2.883942 9.848045 6 | 5.471608 4.896505 1.12 0.265 -4.185609 15.12883 7 | 6.814889 3.479129 1.96 0.052 -.0468846 13.67666 8 | 3.069574 3.77372 0.81 0.417 -4.373212 10.51236 9 | -1.535927 3.107452 -0.49 0.622 -7.664653 4.592799 10 | -2.726272 3.958503 -0.69 0.492 -10.5335 5.080955 11 | .3786996 3.313104 0.11 0.909 -6.155628 6.913027 12 | -.0614476 3.467889 -0.02 0.986 -6.901053 6.778158 | _cons | 43.16242 2.869893 15.04 0.000 37.50223 48.82262 ------------------------------------------------------------------------------ . . **** . * Column 6: Association between Google searches about abortion and 16P broadcast week . reg howgetabort new16p trend trend2 i.month Source | SS df MS Number of obs = 209 -------------+---------------------------------- F(14, 194) = 44.40 Model | 54076.6835 14 3862.62025 Prob > F = 0.0000 Residual | 16877.7663 194 86.9987953 R-squared = 0.7621 -------------+---------------------------------- Adj R-squared = 0.7450 Total | 70954.4498 208 341.127162 Root MSE = 9.3273 ------------------------------------------------------------------------------ howgetabort | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- new16p | -1.62702 1.730145 -0.94 0.348 -5.039328 1.785289 trend | .5597848 .0443047 12.63 0.000 .472404 .6471656 trend2 | -.0014187 .0002045 -6.94 0.000 -.001822 -.0010153 | month | 2 | -2.025201 3.174232 -0.64 0.524 -8.285635 4.235234 3 | 1.709383 3.144418 0.54 0.587 -4.492251 7.911017 4 | -3.914336 3.251354 -1.20 0.230 -10.32688 2.498205 5 | -2.804359 3.111576 -0.90 0.369 -8.94122 3.332502 6 | -6.844806 3.288236 -2.08 0.039 -13.33009 -.3595244 7 | -4.046143 3.096227 -1.31 0.193 -10.15273 2.060445 8 | -10.45961 3.078455 -3.40 0.001 -16.53114 -4.388069 9 | -12.95626 3.139533 -4.13 0.000 -19.14825 -6.764258 10 | -12.02641 3.094 -3.89 0.000 -18.12861 -5.924212 11 | -7.411403 3.154027 -2.35 0.020 -13.63199 -1.190817 12 | -13.70534 3.18155 -4.31 0.000 -19.98021 -7.430473 | _cons | 33.77141 2.770661 12.19 0.000 28.30692 39.23589 ------------------------------------------------------------------------------ . newey howgetabort new16p trend trend2 i.month, lag(1) force Regression with Newey-West standard errors Number of obs = 209 maximum lag: 1 F( 14, 194) = 57.14 Prob > F = 0.0000 ------------------------------------------------------------------------------ | Newey-West howgetabort | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- new16p | -1.62702 2.166873 -0.75 0.454 -5.900673 2.646633 trend | .5597848 .0526781 10.63 0.000 .4558895 .6636801 trend2 | -.0014187 .0002422 -5.86 0.000 -.0018964 -.0009409 | month | 2 | -2.025201 5.281799 -0.38 0.702 -12.44232 8.39192 3 | 1.709383 4.562771 0.37 0.708 -7.289623 10.70839 4 | -3.914336 4.705969 -0.83 0.407 -13.19576 5.367093 5 | -2.804359 4.872921 -0.58 0.566 -12.41506 6.806344 6 | -6.844806 5.57359 -1.23 0.221 -17.83742 4.147804 7 | -4.046143 4.243073 -0.95 0.341 -12.41462 4.322333 8 | -10.45961 4.47483 -2.34 0.020 -19.28517 -1.634043 9 | -12.95626 4.539203 -2.85 0.005 -21.90878 -4.003734 10 | -12.02641 4.164441 -2.89 0.004 -20.2398 -3.813018 11 | -7.411403 4.668748 -1.59 0.114 -16.61942 1.796617 12 | -13.70534 4.620455 -2.97 0.003 -22.81811 -4.592569 | _cons | 33.77141 4.317244 7.82 0.000 25.25665 42.28617 ------------------------------------------------------------------------------ . . **** . * Column 7: Association between Google searches about abortion and Google searches for 16P . * Replicates KL Table 4, Panel A, column 3 . reg howgetabort index16p trend trend2 i.month Source | SS df MS Number of obs = 209 -------------+---------------------------------- F(14, 194) = 45.39 Model | 54357.7195 14 3882.69425 Prob > F = 0.0000 Residual | 16596.7303 194 85.5501562 R-squared = 0.7661 -------------+---------------------------------- Adj R-squared = 0.7492 Total | 70954.4498 208 341.127162 Root MSE = 9.2493 ------------------------------------------------------------------------------ howgetabort | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- index16p | -.0735905 .0359755 -2.05 0.042 -.1445438 -.0026371 trend | .6223141 .055392 11.23 0.000 .5130663 .7315618 trend2 | -.0016883 .00025 -6.75 0.000 -.0021815 -.0011952 | month | 2 | -2.170842 3.138829 -0.69 0.490 -8.361452 4.019769 3 | 1.447416 3.089299 0.47 0.640 -4.645509 7.540341 4 | -3.538073 3.150649 -1.12 0.263 -9.751996 2.67585 5 | -3.377624 3.00339 -1.12 0.262 -9.301113 2.545866 6 | -7.328894 3.148873 -2.33 0.021 -13.53931 -1.118475 7 | -4.296553 3.057529 -1.41 0.162 -10.32682 1.733712 8 | -11.3889 3.092027 -3.68 0.000 -17.4872 -5.290594 9 | -13.87557 3.150285 -4.40 0.000 -20.08877 -7.662362 10 | -12.99011 3.097172 -4.19 0.000 -19.09857 -6.881662 11 | -8.026285 3.115228 -2.58 0.011 -14.17035 -1.882221 12 | -14.03573 3.136063 -4.48 0.000 -20.22088 -7.850572 | _cons | 33.60578 2.742618 12.25 0.000 28.1966 39.01495 ------------------------------------------------------------------------------ . newey howgetabort index16p trend trend2 i.month, lag(1) force Regression with Newey-West standard errors Number of obs = 209 maximum lag: 1 F( 14, 194) = 59.36 Prob > F = 0.0000 ------------------------------------------------------------------------------ | Newey-West howgetabort | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- index16p | -.0735905 .0525843 -1.40 0.163 -.1773007 .0301198 trend | .6223141 .0681078 9.14 0.000 .4879873 .7566408 trend2 | -.0016883 .0003054 -5.53 0.000 -.0022907 -.001086 | month | 2 | -2.170842 5.083271 -0.43 0.670 -12.19641 7.854728 3 | 1.447416 4.45103 0.33 0.745 -7.331205 10.22604 4 | -3.538073 4.590649 -0.77 0.442 -12.59206 5.515914 5 | -3.377624 4.759118 -0.71 0.479 -12.76388 6.008631 6 | -7.328894 5.525035 -1.33 0.186 -18.22574 3.567953 7 | -4.296553 4.129225 -1.04 0.299 -12.44049 3.847383 8 | -11.3889 4.298058 -2.65 0.009 -19.86582 -2.911977 9 | -13.87557 4.384573 -3.16 0.002 -22.52312 -5.228017 10 | -12.99011 3.96543 -3.28 0.001 -20.811 -5.169225 11 | -8.026285 4.554981 -1.76 0.080 -17.00993 .9573562 12 | -14.03573 4.448594 -3.16 0.002 -22.80955 -5.261909 | _cons | 33.60578 4.248878 7.91 0.000 25.22585 41.9857 ------------------------------------------------------------------------------ . . . clear . . ************ . . . ************ . * Data: Replicating State-level Google searches . * Data file obtained from https://www.aeaweb.org/aer/data/10512/20140012_data.zip . * . * See footnote 17 . * . use "$work/data/googletrends-state.dta" . . **** . * Create state categorical variable . quietly encode stname, gen(state) . . **** . * Create 2nd period dummy variable . gen yeardv=(year==200910) . ************ . . . ************ . * Replication State-Level Google Results . . **** . * Association between Google searches for Birth Control and Google searches for 16 and Pregnant > , weighted . reg hgbc index16p yeardv i.state [weight=fempop] (analytic weights assumed) (sum of wgt is 8.4363e+07) Source | SS df MS Number of obs = 30 -------------+---------------------------------- F(16, 13) = 28.31 Model | 6110.8724 16 381.929525 Prob > F = 0.0000 Residual | 175.351762 13 13.4885971 R-squared = 0.9721 -------------+---------------------------------- Adj R-squared = 0.9378 Total | 6286.22417 29 216.766351 Root MSE = 3.6727 ------------------------------------------------------------------------------ hgbc | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- index16p | .7513827 .12722 5.91 0.000 .4765405 1.026225 yeardv | -36.87388 5.681265 -6.49 0.000 -49.14751 -24.60026 | state | FL | 24.30075 2.975054 8.17 0.000 17.87353 30.72796 GA | 17.5794 3.667301 4.79 0.000 9.656683 25.50213 IL | 15.52002 3.235144 4.80 0.000 8.530911 22.50912 MA | 20.23617 4.214491 4.80 0.000 11.13132 29.34103 MI | 17.50653 4.015041 4.36 0.001 8.832566 26.1805 MO | 17.15537 4.452953 3.85 0.002 7.53535 26.77539 NC | 37.87533 3.79714 9.97 0.000 29.67211 46.07855 NJ | 14.72305 3.791193 3.88 0.002 6.532681 22.91343 NY | 14.08424 2.797809 5.03 0.000 8.039945 20.12854 OH | 18.04141 3.725631 4.84 0.000 9.992678 26.09015 PA | 31.41607 3.649424 8.61 0.000 23.53197 39.30017 TX | 18.5649 2.721426 6.82 0.000 12.68562 24.44419 VA | 4.545924 3.762281 1.21 0.248 -3.58199 12.67384 WA | 1.553513 4.151842 0.37 0.714 -7.415997 10.52302 | _cons | 57.18381 1.967408 29.07 0.000 52.93349 61.43414 ------------------------------------------------------------------------------ . . **** . * Association between Google searches for Birth Control and Google searches for 16 and Pregnant > , unweighted . reg hgbc index16p yeardv i.state Source | SS df MS Number of obs = 30 -------------+---------------------------------- F(16, 13) = 13.03 Model | 4940.69187 16 308.793242 Prob > F = 0.0000 Residual | 307.985732 13 23.6912101 R-squared = 0.9413 -------------+---------------------------------- Adj R-squared = 0.8691 Total | 5248.6776 29 180.988883 Root MSE = 4.8674 ------------------------------------------------------------------------------ hgbc | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- index16p | .7347931 .1875573 3.92 0.002 .3296003 1.139986 yeardv | -34.88193 8.781244 -3.97 0.002 -53.85265 -15.91121 | state | FL | 24.44176 5.121797 4.77 0.000 13.37679 35.50673 GA | 17.75152 5.241924 3.39 0.005 6.427034 29.07601 IL | 15.65481 5.100342 3.07 0.009 4.636187 26.67343 MA | 20.40207 5.216223 3.91 0.002 9.133104 31.67103 MI | 17.80307 5.910249 3.01 0.010 5.034756 30.57139 MO | 17.35341 5.357628 3.24 0.006 5.778956 28.92786 NC | 38.05885 5.291146 7.19 0.000 26.62803 49.48968 NJ | 14.85888 5.103857 2.91 0.012 3.832669 25.88509 NY | 14.1983 5.035266 2.82 0.014 3.320266 25.07633 OH | 18.2944 5.64555 3.24 0.006 6.097937 30.49087 PA | 31.67321 5.66945 5.59 0.000 19.42511 43.92131 TX | 18.72354 5.187284 3.61 0.003 7.517096 29.92999 VA | 4.55733 4.869069 0.94 0.366 -5.961655 15.07631 WA | 1.661345 5.017716 0.33 0.746 -9.178771 12.50146 | _cons | 56.41283 3.964702 14.23 0.000 47.84762 64.97805 ------------------------------------------------------------------------------ . . **** . * Association between Google searches for Birth Control and Google searches for 16 and Pregnant > , weighted . reg hgabort index16p yeardv i.state [weight=fempop] (analytic weights assumed) (sum of wgt is 7.5432e+07) Source | SS df MS Number of obs = 24 -------------+---------------------------------- F(13, 10) = 8.35 Model | 4472.87014 13 344.066934 Prob > F = 0.0010 Residual | 412.14148 10 41.214148 R-squared = 0.9156 -------------+---------------------------------- Adj R-squared = 0.8060 Total | 4885.01162 23 212.39181 Root MSE = 6.4198 ------------------------------------------------------------------------------ hgabort | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- index16p | .50459 .2391641 2.11 0.061 -.0283008 1.037481 yeardv | -19.43486 10.60509 -1.83 0.097 -43.06447 4.194749 | state | FL | 28.21098 5.510494 5.12 0.000 15.93284 40.48913 GA | 14.38988 6.792391 2.12 0.060 -.7445114 29.52427 IL | 4.900206 5.989117 0.82 0.432 -8.444378 18.24479 MA | 12.7666 7.800662 1.64 0.133 -4.614359 30.14756 MI | 25.35545 7.461102 3.40 0.007 8.731081 41.97983 NJ | 32.74367 7.015267 4.67 0.001 17.11268 48.37466 NY | 28.28094 5.1791 5.46 0.000 16.74119 39.8207 OH | 16.7425 6.91737 2.42 0.036 1.329641 32.15536 PA | 25.67885 6.77831 3.79 0.004 10.57584 40.78187 TX | 15.36236 5.046635 3.04 0.012 4.117754 26.60696 VA | 1.153094 6.95269 0.17 0.872 -14.33846 16.64465 | _cons | 60.74893 3.651204 16.64 0.000 52.61354 68.88432 ------------------------------------------------------------------------------ . . **** . * Association between Google searches for Birth Control and Google searches for 16 and Pregnant > , unweighted . reg hgabort index16p yeardv i.state Source | SS df MS Number of obs = 24 -------------+---------------------------------- F(13, 10) = 6.23 Model | 3768.7356 13 289.902739 Prob > F = 0.0033 Residual | 465.383539 10 46.5383539 R-squared = 0.8901 -------------+---------------------------------- Adj R-squared = 0.7472 Total | 4234.11914 23 184.092137 Root MSE = 6.8219 ------------------------------------------------------------------------------ hgabort | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- index16p | .3954854 .270143 1.46 0.174 -.2064309 .9974016 yeardv | -15.63365 12.63327 -1.24 0.244 -43.78233 12.51503 | state | FL | 29.13837 7.197983 4.05 0.002 13.10027 45.17648 GA | 15.52184 7.375207 2.10 0.062 -.9111449 31.95482 IL | 5.786682 7.16631 0.81 0.438 -10.18085 21.75421 MA | 13.85765 7.337307 1.89 0.088 -2.490893 30.20619 MI | 27.3057 8.357974 3.27 0.008 8.682973 45.92843 NJ | 33.63696 7.171499 4.69 0.001 17.65787 49.61606 NY | 29.03104 7.070196 4.11 0.002 13.27766 44.78442 OH | 18.40635 7.969324 2.31 0.044 .6495873 36.16311 PA | 27.36998 8.004445 3.42 0.007 9.534962 45.20499 TX | 16.40567 7.294621 2.25 0.048 .1522425 32.6591 VA | 1.228104 6.82443 0.18 0.861 -13.97767 16.43388 | _cons | 60.32805 5.607616 10.76 0.000 47.83351 72.8226 ------------------------------------------------------------------------------ . . . clear . . ************ . . . . . . capture log close