README

This document explains the data sources and how to use the replication materials for "The negative consequences of sports gambling opportunities on human capital formation: evidence from Spain", Espadafor & Martínez (2021).

All the code and data will be also published and available in Harvard Dataverse once the paper is accepted for publication.

This replication package contains Stata/R data- and do-files/scripts to generate the results reported in "The negative consequences of sports gambling opportunities on human capital formation: evidence from Spain" by Espadafor & Martínez (2021).

Please carefully read this document to replicate the resutls of this study. If you face any problem, drop us a line: maria.canizares@eui.eu & sergi.martinez@eui.eu

Data description and sources

This section details the source of the data employed in this study.

This study uses several sources for data, which are publicly available. The only exception is the data from the Millenium Cohort Study, which we use to estimate the proportion of students who gamble in the UK by age 14 (footnote 12 in the Manuscript). MCS data are available through the UK Data Service, see more information in how to access it here:https://cls.ucl.ac.uk/cls-studies/millennium-cohort-study/ The 07_gambling_mcs.do contains information and the code for calculations of Footnote 12.

• Information on schools' characteristics and educational performance come from public administrative data yearly released by Madrid's education authorities, which is available here: https://gestiona3.madrid.org/wpad_pub/run/j/MostrarConsultaGeneral.icm. A version of this data, including educational centers that offered the academic track in post-compulsory education (Bachillerato, in Spanish), has been published and it is publicly available to download here: Espadafor, Mar, 2021, "Panel Data on High Schools in Madrid (2013-2018)", https://doi.org/10.7910/DVN/K4NSBK, Harvard Dataverse, V1, UNF:6:B5vf/YEfEjiSv/71pOi4Kw== [fileUNF]

• Information on openings of betting houses is available as part of the census data provided by the Government of Madrid, which is also accessible here: https://datos.madrid.es/portal/site/egob/menuitem.c05c1f754a33a9fbe4b2e4b284f1a5a0/?vgnextoid=23160329ff639410VgnVCM2000000c205a0aRCRD&vgnextchannel=374512b9ace9f310VgnVCM100000171f5a0aRCRD&vgnextfmt=default Information about the betting houses was extracted by the authors using the licenses issued by the Government of Madrid to betting houses companies. These licenses come from the following census record: "Censo de locales, sus actividades y terrazas de hostelería y restauración". We restricted our sample to private gambling facilities, excluding stores regulated by Loterías y Apuestas del Estado, the State-owned enterprise in charge of the most popular both football and lottery games.

We coded each license's location and the year in which it was issued to match betting houses to high schools. After this, we calculated walking distances in meters from each high school to the closest betting house to assess each high school's access to gambling facilities. We employ a radius to divide Madrid high schools using a dummy variable switching on those education centers exposed to a betting house at less than 500m. The rationale behind this choice is that, according to the "Encuesta Domiciliaria de Mobilidad 2018" (Consorcio de Mobilidad de Madrid, Comunidad de Madrid), the average Madrid students' commuting to educational centers is 500 meters, as suggested by descriptive statistics shown in Table. S.1 and Figure. S.1. in the Supplementary Materials document.

Additionally, we extracted and linked the average household income level of the neighbourhood where high schools are located.

• Information on neighborhoods income levels is extracted from the "Panel de indicadores de distritos y barrios de Madrid. Estudio sociodemográfico" 2016-2020, which is also publicly available here: https://datos.madrid.es/sites/v/index.jsp?vgnextoid=71359583a773a510VgnVCM2000001f4a900aRCRD&vgnextchannel=374512b9ace9f310VgnVCM100000171f5a0aRCRD

As part of the placebo tests, we extracted licenses issued to Starbucks. We then geolocated them and calculated walking distances from each high school to the closest Starbucks.

• Information on licences to Starbucks can be found in the census data provided by the Government of Madrid (see link above).

Finally, as part of the robustness checks, we used monthly rent prices by districts. We linked each high school to the rent prices in their area.

• The data on rent prices is extracted from Madrid open access records, "Renta mensual de la vivienda en alquiler (€/m2 construido) por Distrito y por Trimestre". It can be downloaded here: https://www.madrid.es/portales/munimadrid/es/Inicio/El-Ayuntamiento/Estadistica/Areas-de-informacion-estadistica/Edificacion-y-vivienda/Mercado-de-la-vivienda/Precios-de-la-vivienda/?vgnextfmt=default&vgnextoid=bf281b47a277b210VgnVCM1000000b205a0aRCRD&vgnextchannel=22613c7ea422a210VgnVCM1000000b205a0aRCRD

A copy of all the data is provided as part of this archive. The data are in the public domain.

Replication Folder and Files description

The Replication Material is structured as follows:

• Data folder contains all datasets needed to replicate the analysis presented both in the manuscript and the Supplementary Information
• Scripts folder contains the code to replicate tables and figures
• Estimates folder contains all regression estimates saved as .dta files (these come from the scripts in the folder above)
• Tables folder contains all tables.text outcomes from the regressions in their respective code

To replicate the results of Espadafor & Martínez (2021) you will need the following files and codes listed in the table bellow.

• The file gambling_data.dta includes all the main variables ready to replicate the main analysis of this study.
• The files 01_estimates_main.do to 02_estimates_logdist.do has the code to reproduce results in each figure in the manuscript.
• The file 06_code_for_PanelMatchdata.do has the code to reproduce results using the Callaway and Sant Anna (2020) estimator for staggered difference-in-differences designs, which is reeported in Fig.4 and then in Fig S14-15 in the Supplementary Information Document.
• The file 03_estimates_placebo_number_student.do has the code to reproduce results using a placebo outcome: number of students.

The following files creates additional results under alternative specifications and alternative outcomes as reported in the online Supplementary Material document:

• The file 08_SI.do has the code to reproduce all tables presented in the SI:
  • All summary statistics (Tables S2-3), Average effect and time-placebos (Tables S5-9)
  • The likelihood of school being treated by type of schoool (Table S4)
  • Replication of main results using logdistance (Table S10)
  • Results from robutsness checks for compositonal change using number of students (Tables S11-S14) and rent prices (Tables S15-17)
  • Restuls from placebo test using Starbacks (Table S18)
  • Results using a different treatment definition, <200m (Table S19)

Figure & tables list

Figure/Table	Script	Data	Note:
Figure 1	figure1.r	apuestas_tipos_.dta
Figure 2	figure2.r	pau2.dta	Source: 01_estimates_main.do
Figure 3	figure3.r	distance2.dta	Source: 02_estimates_logdist.do
Figure 4	figure4.r	gambling_data.dta
Figure 5	figure5.r	gambling_data.dta, num2.dta	Source: 03_estimates_placebo_number_student
Figure S1	figureS1.r	movilidad_merged_1503.dta
Figure S2	figureS2.r	apuestas_tipos_.dta, colegiosX_.dta, all_mapa2.dta
Figure S3	figureS3.r
Figure S4	figureS4.r	panelmatch.dta	Source: 06_code_for_PanelMatchdata.do
Figure S5	figureS5.r	pau2.dta	Source: 01_estimates_main.do
Figure S6	figureS6.r	gambling_data.dta
Figure S7	figureS7.r	gambling_data.dta
Figure S8	figureS8.r	distance2.dta	Source: 02_estimates_logdist.do
Figure S9	figureS9.r	gambling_data.dta, num2.dta	Source: 03_estimates_placebo_number_student
Figure S10	figureS10.r	price2.dta	Source: 04_estimates_placebo_rentprices.do
Figure S11	figureS11.r	gambling_data.dta
Figure S12	figureS12.r	star1.dta	Source: 05_estimates_placebo_starbucks.do
Figure S13	figureS11.r	gambling_data.dta
Figure S14	figureS11.r	gambling_data.dta
Figure S15	appendix.do	gambling_data.dta
Table S1	appendix.do	gambling_data.dta
Table S2	appendix.do	gambling_data.dta
Table S3	appendix.do	gambling_data.dta
Table S4	appendix.do	gambling_data.dta
Table S5	appendix.do	gambling_data.dta
Table S6	appendix.do	gambling_data.dta
Table S7	appendix.do	gambling_data.dta
Table S8	appendix.do	gambling_data.dta
Table S9	appendix.do	gambling_data.dta
Table S10	appendix.do	gambling_data.dta
Table S11	appendix.do	gambling_data.dta
Table S12	appendix.do	gambling_data.dta
Table S13	appendix.do	gambling_data.dta
Table S14	appendix.do	gambling_data.dta
Table S15	appendix.do	gambling_data.dta
Table S16	appendix.do	gambling_data.dta
Table S17	appendix.do	gambling_data.dta
Table S18	appendix.do	gambling_data.dta

Variable description

Variable name:	Description
school_id	High School's unique id
year
apuestas_id	Betting houses' unique id
km_to_apuestas_id	Distance from school to closest betting house in km
metres	Distance from school to closest betting house in m
metres_years	Distance from school to closest betting house in m, by year
name	High School's name
district	District where high schools are located
postcode	Postcode where high schools are located
pau	Average high school performance in PAU
pauXXXX	Average high school performance in PAU by academic year
num	Average number of students by high school
nXXXX	Average number of students by high school and academic year
publico	Whether schools are publicly financed
treat	Takes 1 if there is a betting house closer than 500m
treat_year	Takes 1 if there is a betting house closer than 500m, by year
logdist_year	Logged distance from school to closest bettng houses by year
alquiler	Rent prices euros/m2 by district
star_id	Starbucks' unique id
km_to_star_id	Distance from school to closest Starbucks in km
metres_year_star	Distance from school to closest Starbucks in m by year
logdist_year_star	Logged distance from school to closest Starbucks
renta	Average income where schools' are located
renta2	Takes 1 if average income is above the average of Madrid
tf	The first period when a particular observation is treated

Data file	Description
gambling_data.dta	Main dataset containing all variables described above
apuestas_tipos_.dta	Dataset containing type of betting houses
all_mapa2.dta	Dataset containing longtitude and latitude from all units
colegios2_.dta	Dataset containing longitude and latitude for all schools
colegios3_.dta	Dataset containing longitude and latitude for all public schools
colegios4_.dta	Dataset containing longitude and latitude for all private schools
dapuestas_year.dta	Dataset containing betting house information by year

Software Requirements

• Stata (code was last run with version 16)

  • estout
  • xtreg
  • estab

• R (code was last run wth version 4.0.4 (2021-02-15))

  • Platform: x86_64-apple-darwin17.0 (64-bit)

  • Running under: macOS Mojave 10.14.3

  • Attached base packages: [1] stats graphics grDevices utils datasets methods base

  • other attached packages: [1] foreign_0.8-81 sf_0.9-6 sp_1.4-2 ggmap_3.0.0
    [5] ggplot2_3.3.3 haven_2.3.1

  • loaded via a namespace (and not attached): [1] Rcpp_1.0.6 pillar_1.5.0 compiler_4.0.4
    [4] plyr_1.8.6 class_7.3-18 forcats_0.5.1
    [7] bitops_1.0-6 tools_4.0.4 lifecycle_1.0.0
    [10] tibble_3.0.6 gtable_0.3.0 lattice_0.20-41
    [13] pkgconfig_2.0.3 png_0.1-7 rlang_0.4.10
    [16] cli_2.3.0 DBI_1.1.1 e1071_1.7-3
    [19] stringr_1.4.0 withr_2.4.1 dplyr_1.0.4
    [22] httr_1.4.2 generics_0.1.0 RgoogleMaps_1.4.5.3 [25] vctrs_0.3.6 hms_1.0.0 classInt_0.4-3
    [28] grid_4.0.4 tidyselect_1.1.0 glue_1.4.2
    [31] R6_2.5.0 jpeg_0.1-8.1 fansi_0.4.2
    [34] readr_1.4.0 tidyr_1.1.2 purrr_0.3.4
    [37] magrittr_2.0.1 units_0.6-7 scales_1.1.1
    [40] ellipsis_0.3.1 assertthat_0.2.1 colorspace_2.0-0
    [43] KernSmooth_2.23-18 utf8_1.1.4 stringi_1.5.3
    [46] munsell_0.5.0 rjson_0.2.20 crayon_1.4.1