EMPIRICAL NOTES

£ il :і i.

Tables

Table 1.1 Health and demographic characteristics of insured and uninsured couples in the

NHIS

Data source. The 2009 NHIS data are from the Integrated Health Interview Series (IHIS) and are available at www. ihis. us/ihis/.

Sample. The sample used to construct this table consists of husbands and wives aged 26- 59, with at least one spouse working.

Variable definitions. Insurance status is determined by the IHIS variable UNINSURED. The health index is on a five-point scale, where 1 = poor, 2 = fair, 3 = good, 4 = very good, 5 = excellent; this comes from the variable HEALTH. Education is constructed from the variable EDUC and measures completed years of schooling. High school graduates and GED holders are assigned 12 years of schooling. People with some college but no degree, and those with an associate’s degree, are assigned 14 years of schooling. Bachelor’s degree holders are assigned 16 years of schooling, and holders of higher degrees are assigned 18 years of schooling. Employed individuals are those “working for pay” or “with job but not at work” as indicated by the variable EMPSTAT.

Family income is constructed by assigning to each bracket of the IHIS income variable (INCFAM07ON) the average household income for that bracket based on data from the 2010 Current Population Survey (CPS) March supplement (using the CPS variable FTOTVAL). The CPS sample used for this purpose omits observations with nonpositive household income as well as observations with negative weights. CPS income is censored at the 98th percentile; values above the 98th percentile are assigned 1.5 times the 98th percentile value.

Additional table notes. A11 calculations are weighted using the variable PERWEIGHT. Robust standard errors are shown in parentheses.

Table 1.3 Demographic characteristics and baseline health in the RAND HIE

Data source. The RAND HIE data are from Joseph P. Newhouse, “RAND Health Insurance Experiment [in Metropolitan and Non-Metropolitan Areas of the United States], 1974-1982,” ICPSR06439-vl, Inter-University Consortium for Political and Social Research, 1999. This data set is available at http://doi. org/10.3886/ICPSR06439.vl.

Sample. The sample used to construct this table consists of adult participants (14 years old and older) with valid enrollment, expenditure, and study exit data.

Variable definitions. The demographic variables in panel A and the health characteristics in panel В are measured at the experimental baseline. The general health index rates the participant’s perception of his or her general health at the time of enrollment. Higher values indicate more favorable self-ratings of health; less health-related worry; and greater perceived resistance to illness. The mental health index rates the participant’s mental health, combining measures of anxiety, depression, and psychological well-being. Higher values indicate better mental health. The education variable measures number of years of completed education and is only defined for individuals 16 years and older. Family income is in constant 1991 dollars.

Additional table notes. Standard errors in parentheses are clustered at the family level.

Table 1.4 Health expenditure and health outcomes in the RAND HIE

Data source. See note for Table 1.3.

Sample. See note for Table 1.3. The panel A sample contains multiple observations for the same person from a different follow-up year.

Variable definitions. See notes for Table 1.3. Variables in panel A are constructed from administrative claims data for each year, and variables in panel В are measured upon exit from the experiment. Face-to-face visits counts the number of face-to-face visits with health professionals that were covered by insurance (excluding dental, psychotherapy, and radiology/anaesthesiology/pathology-only visits). Hospital admissions indicates the total number of covered participant hospitalizations, including admissions for reasons of mental health. The expenditure variables are in constant 1991 dollars.

Additional table notes. Standard errors in parentheses are clustered at the family level.

Table 1.5 OHP effects on insurance coverage and health-care use

Sources. The numbers in columns (1) and (2) are from Amy N. Finkelstein et al., “The Oregon Health Insurance Experiment: Evidence from the First Year,” Quarterly Journal of Economics, vol. 127, no. 3, August 2012, pages 1057-1106. Our numbers come from the original as follows:

■ row (1) in panel A from row (1), columns (1) and (2) in Table III;

■ row (2) in panel A from row (1), columns (1) and (2) in Table IV;

■ row (1) in panel В from row (2), columns (5) and (6) in Table V; and

■ row (2) in panel В from row (1), columns (1) and (2) in Table V.

The numbers reported in columns (3) and (4) are from Sarah L. Taubman et al., “Medicaid Increases Emergency-Department Use: Evidence from Oregon’s Health Insurance Experiment,” Science, vol. 343, no. 6168, January 17, 2014, pages 263-268. Our numbers come from the original as follows:

■ row (1) from row (1), columns (1) and (2) in Table S7;

■ row (3) from row (1), columns (3) and (4) in Table S2;

■ row (4) from row (1), columns (7) and (8) in Table S2.

Samples. Columns (1) and (2) in panel A use the full sample analyzed in the hospital discharge and mortality data in Finkelstein et al. (2012). Columns (3) and (4) in panel A are drawn from the emergency department records of 12 Portland area emergency departments for visits occurring between March 10, 2008 and September 30, 2009. Panel В uses the follow-up survey data analyzed in Finkelstein et al. (2012).

Variable definitions. The variable in row (1) in panel A is a dummy for Medicaid enrollment in the study period (from lottery notification through the end of September 2009), obtained from Medicaid administrative data. The variable in row (2) in panel A is a dummy equal to 1 if the respondent had a non-childbirth hospitalization from notification until the end of August 2009. The variables in rows (3) and (4) in panel A indicate any emergency department visit and count the number of such visits. The variable in row (1) in panel В measures the number of non-childbirth-related outpatient visits in the past 6 months. The variable in row (2) in panel В is a dummy for whether the patient had a prescription drug at the time of the survey.

Additional table notes. Standard errors in parentheses are clustered at the household level.

Table 1.6 OHP effects on health indicators and financial health

Sources. See notes for Table 1.5. The numbers in row (1) in panel A in this table are obtained from row (2), columns (1) and (2) in Table IX in Finkelstein et al. (2012). The numbers reported in columns (3) and (4) are from Katherine Baicker et al., “The Oregon Experiment—Effects of Medicaid on Clinical Outcomes,” New England Journal of Medicine, vol. 368, no. 18, May 2, 2013, pages 1713-1722.

The numbers in columns (3) and (4) come from columns (1) and (2) in the original as follows:

■ row (2) in panel A from row (3) in Table S2;

■ row (3) in panel A from row (2) in Table S2;

■ row (4) in panel A from row (6) in Table SI;

■ row (5) in panel A from row (1) in Table SI;

■ row (1) in panel В from row (3) in Table S3; and

■ row (2) in panel В from row (4) in Table S3.

We thank Amy Finkelstein and Allyson Barnett for providing unpublished standard errors for estimates from Baicker et al. (2013).

Samples. Columns (1) and (2) use the sample from the (first) follow-up survey analyzed in Finkelstein et al. (2012). Columns (3) and (4) use the sample from the (second) follow-up survey analyzed in Baicker et al. (2013).

Variable definitions. The variable in row (1) in panel A is a dummy for whether the respondent rated his or her health as good, very good, or excellent (as compared to fair or poor). Rows (2) and (3) in panel A contain the SF-8 physical and mental component scores. Higher SF-8 scores indicate better health. The scale is normalized to have a mean of 50 and standard deviation of 10 in the U. S. population; the range is 0 to 100. See pages 14-16 of the appendix of Baicker et al. (2013) for descriptions of the subjective and clinical measures of health used in rows (2)-(5). The variable in row (1) in panel В is a dummy for whether health expenditures surpassed 30% of total income in the past 12 months. The variable in row (2) in panel В is a dummy for whether the respondent had any medical debt at the time of the survey.

Additional table notes. Standard errors in parentheses are clustered at the household

level.

Table 2.2 Private school effects: Barron’s matches

Data sources. The data used to construct this table are described in Stacy Berg Dale and Alan B. Krueger, “Estimating the Payoff to Attending a More Selective College: An Application of Selection on Observables and Unobservables,” Quarterly Journal of Economics, vol. 117, no. 4, November 2002, pages 1491-1527.

These data are from the College and Beyond (C&B) survey linked to a survey administered by Mathematica Policy Research, Inc., in 1995-1997 and to files provided by the College Entrance Examination Board and the Higher Education Research Institute (HERI) at the University of California, Los Angeles. The college selectivity category is as determined by Barron’s Profiles of American Colleges 1978, Barron’s Educational Series, 1978.

Sample. The sample consists of people from the 1976 college entering cohort who appear in the C&B survey and who were full-time workers in 1995. The analysis excludes students from historically black universities (Howard University, Morehouse College, Spellman College, and Xavier University; see pages 1500-1501 in Dale and Krueger (2002) for details). The sample is further restricted to applicant selectivity groups containing some students who attended public universities and some students who attended private universities.

Variable definitions. The dependent variable is the log of pretax annual earnings in 1995. The question in the C&B survey has 10 income brackets; see footnote 8 on pages 1501-1502 in Dale and Krueger (2002) for exact construction of the earnings variable. The applicant group variable is formed by matching students according to the list of categories of schools where they applied and were accepted or rejected (from the C&B survey), where school categories are based on the Barron’s college selectivity measure (see pages 1502-1503 in Dale and Krueger (2002) for more on this). The variable own SAT score/100 measures the respondent’s SAT score divided by 100. See page 1508 in Dale and Krueger (2002) for the definition of the parental income variable (this is imputed using parental occupation and schooling). Variables female, black, Hispanic, Asian, other/missing race, high school top 10%, high school rank missing, and athlete are dummies.

Additional table notes. Regressions are weighted to make the sample representative of the population of students at C&B institutions (see page 1501 in Dale and Krueger (2002) for details). Standard errors in parentheses are clustered at the level of school attended.

Table 2.3 Private school effects: Average SAT score controls

Data sources. See notes for Table 2.2.

Sample. See notes for Table 2.2. The sample used to construct this table contains all C&B students and not just those with Barron’s selectivity group matches.

Variable definitions. See notes for Table 2.2. The variable average SAT score of schools applied to/100 is constructed as follows: the average SAT score (divided by 100) is computed for each university using HERI data and then averaged over the universities

where each respondent applied.

Additional table notes. Regressions are weighted to make the sample representative of the population of students at C&B institutions. Standard errors in parentheses are clustered at the university level.

Table 2.4 School selectivity effects: Average SAT score controls

Data sources. See notes for Table 2.2.

Sample. See notes for Table 2.3.

Variable definitions. See notes for Table 2.3. The variable school average SAT score/100 is the average SAT score (divided by 100) of the students at the school the respondent attended.

Additional table notes. See notes for Table 2.3.

Table 2.5 Private school effects: Omitted variables bias

Data sources. See notes for Table 2.2.

Sample, variable definitions, and additional table notes. See notes for Table 2.3.

Table 3.1 Analysis of KIPP lotteries

Data sources. Demographic information on students in Lynn public schools is from the Massachusetts Student Information Management System. Demographic and lottery information for KIPP applicants is from KIPP Lynn school records. Scores are from the Massachusetts Comprehensive Assessment System (MCAS) tests in math and English language arts. For details, see Joshua D. Angrist et al., “Who Benefits from KIPP?” Journal of Policy Analysis and Management, vol. 31, no. 4, Fall 2012, pages 837-860.

Sample. The sample in column (1) contains students who attended fifth grade in Fynn public schools between fall 2005 and spring 2008. The samples in columns (2)-(5) are drawn from the set of KIPP Fynn applicants for fifth – and sixth-grade entry in the same period. Applicants with siblings already enrolled in KIPP or who went directly onto the waiting list are excluded (see footnote 14 in Angrist et al. (2012)). Fottery comparisons are limited to the 371 applicants with follow-up data.

Variable definitions. Hispanic, black, female, free/reduced-price lunch, and enrolled at KIPP are dummy variables. The math and verbal scores for students in a given grade are standardized with respect to the reference population of all students in Massachusetts in that grade. Baseline scores are from fourth-grade tests. Outcome scores are from the grades following the application grade, specifically, fifth-grade scores for those who applied to KIPP when they were in fourth grade and sixth grade scores for those who applied to KIPP while in fifth.

Additional table notes. Robust standard errors are reported in parentheses.

Table 3.3 Assigned and delivered treatments in the MDVE

Data sources. The numbers reported in this table are from Table 1 in Fawrence W. Sherman and Richard A. Berk, “The Specific Deterrent Effects of Arrest for Domestic

Assault,” American Sociological Review, vol. 49, no. 2, April 1984, pages 261-272.

Table 3.4 Quantity-quality first stages

Data sources. The data used to construct this table are from the 20% public-use microdata samples from the 1983 and 1995 Israeli Censuses, linked with nonpublic information on parents and siblings from the population registry. For details, see Joshua D. Angrist, Victor Lavy, and Analia Schlosser, “Multiple Experiments for the Causal Link between the Quantity and Quality of Children,” Journal of Labor Economics, vol. 28, no. 4, October 2010, pages 773-824.

Sample. The sample includes Jewish, first-born non-twins aged 18-60. The sample is restricted to individuals whose mothers were born after 1930 and who had their first birth between the ages of 15 and 45.

Variable definitions. The twins instrument (second-born twins) is a dummy variable equal to 1 in families where the second birth produces twins. The sex-mix instrument (same sex) is a dummy variable equal to 1 if the second and first born are same-sex.

Additional table notes. In addition to a dummy for males, additional covariates are dummies for census year, parents’ ethnicities (Asian or African origin, from the former Soviet Union, from Europe or America), and missing month of birth; age, mother’s age, mother’s age at first birth, and mother’s age at immigration (where relevant). The first stages in this table go with the second-stage estimates in the first two rows of Table 3.5. Robust standard errors are reported in parentheses.

Table 3.5 OLS and 2SLS estimates of the quantity-quality trade-off

Data sources. See notes for Table 3.4.

Sample. See notes for Table 3.4. Estimates in the third and fourth rows of the table are limited to subjects aged 24-60 at the time of the census. The college graduation outcome has a few additional missing values.

Variable definitions. See notes for Table 3.4. The dependent variables in the second, third, and fourth rows are dummy variables.

Additional table notes. Covariates are listed in the notes for Table 3.4.

Table 4.1 Sharp RD estimates of ML DA effects on mortality

Data sources. Mortality data are from the National Center for Health Statistics (NCHS) confidential mortality detail files for 1997-2004. These data are derived from death certificates and cover all deaths in the United States in the study period. Population estimates in the denominator are from the 1970-1990 U. S. Censuses. For details, see pages 166-169 of Christopher Carpenter and Carlos Dobkin, “The Effect of Alcohol Consumption on Mortality: Regression Discontinuity Evidence from the Minimum Drinking Age,” American Economic Journal—Applied Economics, vol. 1, no. 1, January 2009, pages 164-182.

Sample. The sample is restricted to fatalities of young adults aged 19-22. The data used here consist of averages in 48 cells defined by age in 30-day intervals.

Variable definitions. Cause of death is reported on death certificates in the NCHS data.

Causes are divided into internal and external, with the latter split into mutually exclusive subcategories: homicide, suicide, motor vehicle accidents, and other external causes. A separate category for alcohol-related causes covers all deaths for which alcohol was mentioned on the death certificate. Outcomes are mortality rates per 100,000, where the denominator comes from census population estimates.

Additional table notes. Robust standard errors are reported in parentheses.

Table 5.1 Wholesale firm failures and sales in 1929 and 1933

Source. Numbers in this table are from Table 8 (page 1066) in Gary Richardson and William Troost, “Monetary Intervention Mitigated Banking Panics during the Great Depression: Quasi-Experimental Evidence from a Federal Reserve District Border, 1929-1933,” Journal of Political Economy, vol. 117, no. 6, December 2009, pages 1031-1073.

Data sources. Data are from the 1935 Census of American Business, as compiled by Richardson and Troost (2009).

Table 5.2 Regression DD estimates of MLDA effects on death rates

Data sources. MLDA provisions by state and year are from “Minimum Purchase Age by State and Beverage, 1933-Present,” DISCUS (Distilled Spirits Council of the US), 1996; Alexander C. Wagenaar, “Legal Minimum Drinking Age Changes in the United States: 1970-1981,” Alcohol Health and Research World, vol. 6, no. 2, Winter 1981- 1982, pages 21-26; and William Du Mouchel, Allan F. Williams, and Paul Zador, “Raising the Alcohol Purchase Age: Its Effects on Fatal Motor Vehicle Crashes in Twenty-Six States,” Journal of Legal Studies, vol. 16, no. 1, January 1987, pages 249- 266. We follow the coding of these laws implemented in Karen E. Norberg, Laura J. Bierut, and Richard A. Grucza, “Long-Term Effects of Minimum Drinking Age Laws on Past-Year Alcohol and Drug Use Disorders,” Alcoholism: Clinical and Experimental Research, vol. 33, no. 12, September 2009, pages 2180-2190, correcting minor coding errors.

Mortality information comes from the Multiple Cause-of-Death Mortality Data available from the National Vital Statistics System of the National Center for Health Statistics, obtained from www. nber. org/data/mortality-data. html. Population data are from the U. S. Census Bureau’s intercensal population estimates available online. See:

http://www. census. gov/popest/data/state/asrh/pre- 1980/tables/e7080sta. txt:

http://www. census. gov/popest/data/state/asrh/1980s/80s st age sex. html: and

http://www. census. gov/popest/data/state/asrh/1990s/st age sex. html.

Sample. The data set used to construct these estimates contains death rates of 18-20-year – olds between 1970 and 1983 by state and year.

Variable definitions. The mortality rate measures the number of 18-20-year-olds who died in a given state and year (per 100,000), by cause of death (all deaths, motor vehicle accidents, suicide, and all internal causes). The MLDA regressor measures the fraction of 18-20-year-olds who are legal drinkers in a given state and year. This fraction is calculated using MLDA change dates in each state and accounts for grandfathering clauses. The calculation assumes that births are distributed uniformly throughout the year.

Additional table notes. Regressions in columns (3) and (4) are weighted by state population aged 18-20. Standard errors in parentheses are clustered at the state level.

Table 5.3 Regression DD estimates of MLDA effects controlling for beer taxes

Data sources. See notes for Table 5.2. Beer tax data are from Norberg et al., “Long-Term Effects,” Alcoholism: Clinical and Experimental Research, 2009.

Sample. See notes for Table 5.2.

Variable definitions. See notes for Table 5.2. The beer tax is measured in constant 1982 dollars per gallon.

Additional table notes. See notes for Table 5.2.

Table 6.2 Returns to schooling for Twinsburg twins

Data sources. The twins data are detailed in Orley Ashenfelter and Cecilia Rouse, “Income, Schooling, and Ability: Evidence from a New Sample of Identical Twins,” Quarterly Journal of Economics, vol. 113, no. 1, February 1998, pages 253-284. These data are available at

http://dataspace. princeton. edu/jspui/handle/88435/dsp01xg94hp567. This includes data used in Orley Ashenfelter and Alan B. Krueger, “Estimates of the Economic Returns to Schooling from a New Sample of Twins,” American Economic Review, vol. 84, no. 5, December 1994, pages 1157-1173.

Sample. The sample consists of 680 twins who were interviewed at the Twinsburg Twins Festival in 1991, 1992, and 1993. The sample is restricted to U. S.-resident twins who have been employed in the 2 years preceding the interview.

Variable definitions. Estimates in this table were constructed using self-reported years of education and sibling reports, defined as an individual’s report of the number of years of education attained by his or her twin sibling.

Additional table notes. Robust standard errors are reported in parentheses.

Table 6.3 Returns to schooling using child labor law instruments

Data sources. The data used to construct this table are detailed in Daron Acemoglu and Joshua D. Angrist, “How Large Are Human-Capital Externalities? Evidence from Compulsory-Schooling Laws,” in Ben S. Bernanke and Kenneth Rogoff (editors), NBER Macroeconomics Annual 2000, vol. 15, MIT Press, 2001, pages 9-59.

Sample. The sample consists of U. S.-born white men aged 40-49, interviewed in U. S. censuses from 1950 through 1990. The sample was drawn from the integrated public use micro data samples (IPUMS) for these censuses.

Variable definitions. The dependent variable is the log weekly wage. The schooling variable is top-coded at 17. The 1990 Census schooling variable is partly imputed using categorical means from other sources. The child labor law instruments are dummies indicating the schooling required before work was allowed in the respondent’s state of birth, according to laws in place at the time the respondent was 14 years old. For details, see pages 22-28 and Appendix В in Acemoglu and Angrist (2001).

Additional table notes. All regressions are weighted using the IPUMS weighting variable. Standard errors in parentheses are clustered at the state level.

Table 6.4 IV recipe for an estimate of the returns to schooling using a single quarter of

birth instrument

Data sources. The data used to construct this table are detailed in Joshua D. Angrist and Alan B. Krueger, “Does Compulsory School Attendance Affect Schooling and Earnings?” Quarterly Journal of Economics, vol. 106, no. 4, November 1991, pages 979-1014.

Sample. The sample consists of men born between 1930 and 1939 in the 1980 U. S. Census 5% public use sample. Observations with allocated values were excluded from the analysis, as were respondents who reported no wage income or no weeks worked in 1979. See pages 1011-1012 in Appendix 1 in Angrist and Krueger (1991).

Variable definitions. Log weekly wages in 1979 are computed by dividing annual earnings by weeks worked. The schooling variable is the highest grade completed.

Additional table notes. Robust standard errors are reported in parentheses.

Table 6.5 Returns to schooling using alternative quarter of birth instruments

Data sources, sample, variable definitions, and additional table notes. See notes for Table 6.4.

Figures

Figure 2.1 The CEF and the regression line

Source. This is Figure 3.1.2 on page 39 in Joshua D. Angrist and Jorn-Steffen Pischke, Mostly Harmless Econometrics: An Empiricist’s Companion, Princeton University Press, 2009.

Sample. See notes for Table 6.4.

Variable definitions. The dependent variable is the log weekly wage. The schooling variable is the highest grade completed.

Figure 3.1 Application and enrollment data from KIPP Lynn lotteries

Data sources. See notes for Table 3.1.

Sample. The KIPP data set analyzed here contains first-time applicants for fifth – and sixth-grade seats in 2005-2008. This sample contains 446 applicants and includes some applicants without follow-up data.

Figure 3.2 IV in school: the effect of KIPP attendance on math scores

Data sources. See notes for Table 3.1.

Sample. The sample here matches that in column (3) of Table 3.1.

Figure 4.1 Birthdays and funerals

Source. This figure is from Appendix A of Christopher Carpenter and Carlos Dobkin, “The Effect of Alcohol Consumption on Mortality: Regression Discontinuity Evidence from the Minimum Drinking Age,” American Economic Journal—Applied Economics, vol. 1, no. 1, January 2009, pages 164-182.

Additional figure notes. The figure plots the number of deaths in the United States between 1997 and 2003 by age in days measured relative to birthdays.

Figure 4.2 A sharp RD estimate of MLDA mortality effects

Data sources and sample. See notes for Table 4.1.

Variable definitions. See notes for Table 4.1. The Y-axis measures mortality (per 100,000) from all causes. Averages in the figure are for 48 cells defined by age in 30- day intervals.

Figure 4.4 Quadratic control in an RD design

Data sources, sample, and variable definitions. See notes for Table 4.1.

Additional figure notes. See notes for Figure 4.2.

Figure 4.5 RD estimates of MLDA effects on mortality by cause of death

Data sources and sample. See notes for Table 4.1.

Variable definitions. See notes for Table 4.1. The Y-axis measures mortality rates per 100,000 population by cause of death. These are averages for 48 cells defined by age in 30-day intervals.

Additional figure notes. See notes for Figure 4.2.

Figure 4.6 Enrollment at BLS

Data sources. This figure uses Boston Public Schools (BPS) data on exam school applications, including information on Independent School Entrance Exam (ISEE) scores, school enrollment status between 1999 and 2008, and MCAS scores from school years 1999/2000 through 2008/2009. For details, see pages 142-143 and appendix C in the supplement to Atila Abdulkadiroglu, Joshua D. Angrist, and Parag Pathak, “The Elite Illusion: Achievement Effects at Boston and New York Exam Schools,” Econometrica, vol. 81, no. 1, January 2014, pages 137-196. The supplement is available at

http://www. econometricsociety. org/ecta/supmat/10266 data description. pdf.

Sample. The sample includes BPS-enrolled students who applied to Boston Latin School (BLS) for seventh grade seats from 1999 to 2008. The sample is restricted to students for whom BLS is either a first choice or a top choice after eliminating schools where the student didn’t qualify.

Variable definitions. The running variable, labeled “entrance exam score” in the figure, is a weighted average of applicants’ ISEE total score and GPA. Exam school enrollment is measured using data from the school year following application.

Additional figure notes. Running variable values in the figure were normalized by subtracting the lowest score offered a seat at BLS in a given year, so that the cutoff for each year is 0. The smoothed lines in the figures are fitted values from regression models estimated with data near each point. These models regress the dependent variable on the running variable for observations with values inside a nonparametric bandwidth. See Abdulkadiroglu et al. (2014) for details.

Figure 4.7 Enrollment at any Boston exam school

Data sources, sample, and additional figure notes. See notes for Figure 4.6.

Variable definitions. See notes for Figure 4.6. Enrollment at any exam school indicates whether an applicant enrolled at Boston Fatin School, Boston Fatin Academy, or the John D. O’Bryant High School of Mathematics and Science.

Figure 4.8 Peer quality around the BFS cutoff

Data sources, sample, and additional figure notes. See notes for Figure 4.6.

Variable definitions. See notes for Figure 4.6. For each exam school applicant, peer quality is the average of the fourth-grade MCAS math scores of his or her schoolmates in seventh grade, at any school he or she attended in that grade.

Figure 4.9 Math scores around the BFS cutoff

Data sources, sample, and additional figure notes. See notes for Figure 4.6.

Variable definitions. See notes for Figure 4.6. The variable on the Y-axis here is the average of seventh – and eighth-grade MCAS math scores.

Figure 4.10 Thistlethwaite and Campbell’s Visual RD

Source. This is Figure 3 in Donald F. Thistlethwaite and Donald T. Campbell, “Regression-Discontinuity Analysis: An Alternative to the ex post facto Experiment,” Journal of Educational Psychology, vol. 51, no. 6, December 1960, pages 309-317.

Sample. The sample contains 5,126 near winners and 2,848 near losers of a Certificate of Merit in the 1957 National Merit Scholarship competition. The running variable is the score on the College Entrance Examination Board’s Scholarship Qualifying Test, now known as the PS AT. The two outcome measures come from a survey administered to all students in the sample approximately 6 months after awards were announced.

Variable definitions. The two outcome variables are dummies for whether a student plans to do 3 or more years of graduate study (plotted as line І-Г), and whether a student plans to be a college teacher or a scientific researcher (plotted as line J-J’).

Figure 5.1 Bank failures in the Sixth and Eighth Federal Reserve Districts

Data sources. Daily data on the number of banks operating in Mississippi were compiled by Gary Richardson and William Troost and are described on pages 1034-1038 of Gary Richardson and William Troost, “Monetary Intervention Mitigated Banking Panics during the Great Depression: Quasi-Experimental Evidence from a Federal Reserve District Border, 1929-1933,” Journal of Political Economy, vol. 117, no. 6, December 2009, pages 1031-1073.

Sample. The bank operations data count all national and state chartered banks in Mississippi, summed within Federal Reserve Districts and in operation on July 1, 1930, and July 1, 1931.

Variable definitions. The Y-axis shows the number of banks open for business on July 1 of a given year in a given district.

Figure 5.2 Trends in bank failures in the Sixth and Eighth Federal Reserve Districts

Data sources. See notes for Figure 5.1.

Sample. The bank operations data count all national and state chartered banks in Mississippi, summed within Federal Reserve Districts, in operation between July 1929 and July 1934.

Variable definitions. See notes for Figure 5.1.

Figure 5.3 Trends in bank failures in the Sixth and Eighth Federal Reserve Districts, and

the Sixth District’s DD counterfactual

Data sources and variable definitions. See notes for Figure 5.1. Sample. See notes for Figure 5.2.

Figure 5.7 John Snow’s DD recipe

Source. This is Table XII (on page 90) in John Snow, On the Mode of Communication of Cholera, second edition, John Churchill, 1855.

Figure 6.1 The quarter of birth first stage

Data sources, sample, and variable definitions. See notes for Table 6.4.

Figure 6.2 The quarter of birth reduced form

Data sources, sample, and variable definitions. See notes for Table 6.4.

Figure 6.3 Last-chance exam scores and Texas sheepskin

Data sources. This figure was constructed using a data set linking administrative high school records, administrative post-secondary schooling records, and unemployment insurance earnings records from Texas. These data are detailed on pages 288-289 of Damon Clark and Paco Martorell, “The Signaling Value of a High School Diploma,” Journal of Political Economy, vol. 122, no. 2, April 2014, pages 282-318.

Sample. The sample consists of five cohorts of seniors taking their last-chance high school exit exam in spring 1993-1997. Earnings data are available through 2004, namely, for a period running from 7 to 11 years after the time of the last-chance exam.

Variable definitions. The running variable on the X-axis measures the score on the last- chance exam, centered around the passing score. Because the exit exam tests multiple subjects and students must pass all to graduate, scores are normalized relative to passing thresholds and the running variable is given by the minimum of these normalized scores. The Y axis plots the probability of diploma receipt conditional on each score value.

Figure 6.4 The effect of last-chance exam scores on earnings

Data sources and sample. See notes for Figure 6.3.

Variable definitions. The running variable on the X-axis is as in Figure 6.3. The Y-axis measures average annual earnings including zeros for those not working conditional on each score value.

[1] Barron’s classifies colleges as Most Competitive, Highly Competitive, Very Competitive, Competitive, Less

[2] A comparison of parametric and nonparametric estimates appears in Tables 4 and 5 of Carpenter and Dobkin, “The

Leave a reply

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>