Working Papers
“Does strengthening HIV criminalization laws affect HIV/AIDS incidence and prevalence?”
I examine the consequences of enhancing the penalties for people who are HIV-positive and expose others to the infection. Using state-year variation in HIV criminalization laws, difference-in-differences models reveal decreases in diagnoses of new HIV infections and in prevalence of HIV. There is no evidence that enhancements in HIV criminalization laws lead to changes in new AIDS cases, further supported by event study analyses that examine the possibility of delayed policy effects. Exploring within-law variation in the type of penalty (misdemeanor, offense, felony) does not reveal heterogeneous effects, nor does examining heterogeneous effects for sexual minorities and non-sexual minorities. Given these significant effects only for HIV, I use individual-level information to explore changes in HIV testing behaviors and in use of HIV-prevention services as potential mechanisms. The results support that the laws resulted in more frequent testing and use of preventive sexual health medical care and a decrease in the number of sexual partners.
“Anti-bullying legislation and academic performance across OECD countries”
Despite the limited number of studies evaluating the impact of bullying behaviors on schooling, there are no studies on whether such policy interventions directly affect schooling, including performance in standardized math and reading tests and high school and college graduation rates. In addition, no studies evaluate the impact of these policies on multiple countries. I start with difference-in-differences models to show that anti-bullying laws decrease the share of bullying victimization and bullying perpetration across countries. However, there exists heterogeneity based on social attitudes towards victimization. The rest of the paper evaluates whether anti-bullying laws directly affect schooling for these two groups of countries using the same difference-in-differences framework. Event study analyses provide evidence that the laws have an effect on improving educational outcomes, though these effects are stronger for countries where other victimization-prevention policies are lacking.
“The bullying-education gradient: Victims and perpetrators of bullying behaviors in developed and developing countries”
Prior research has uncovered an adverse effect of bullying on academic performance in a few developed countries. However, the evidence is still elusive for most developed countries and non-existent for developing countries. The goal of the paper is twofold. First, I examine what determines the probability of being the victim and the perpetrator of bullying across these two groups of countries. However, because students in international databases are not followed across different years and because they do not include detailed information on academic performance, I aggregate the individual-level data to create (and track) groups of students based on their gender, cohort, and country of residence. In the second part of the analysis, I focus on the direct relationship between bullying rates and schooling. I use a regression discontinuity design to examine whether country-level bullying laws significantly affect bullying rates and whether such laws have differential effects based on the level of economic growth. Then, using compulsory bullying laws as instruments for bullying, I follow synthetic cohorts from different countries to estimate the impact of bullying rates on educational attainment. This paper will allow us to evaluate whether bullying prevalence is detrimental for certain regions across the world and whether economic growth is hindered by bullying victimization by limiting the accumulation of human capital.
“The effect of marriage equality on bullying victimization of sexual minority youth”
Legal inequalities may increase hate crimes because they provide opportunities for discrimination and violence. Although several states introduced state policies recognizing same-sex marriage before 2015, there are still hate crimes against this population group. Among youth, hate crimes may be realized in the form of in-school victimization, with some evidence that the adverse effects may be even more exaggerated among LGBTQ students. However, due to data limitations, there is limited evidence on the effect of bullying on youth outcomes, and it is even harder to assess whether youth of different sexual orientations face different probabilities of being bullied. In this paper, I examine whether public policies toward marriage equality for gay and lesbian rights affect the incidence of bullying victimization based on sexual orientation. Using panel data from 2009 to 2019, I analyze whether state-specific policies for legalization of same-sex marriage before the 2015 federal recognition affect such victimization. I also investigate the degree of tolerance towards LGBTQ individuals—as captured by the amount of other recognized LGBTQ rights (e.g., right to adoption, anti-discrimination laws)—that may affect the incidence of bullying victimization.
“Gay, obese, and bullied? Heterogeneous effects of bullying and cyberbullying on health behaviors”
Even though in the 2000s various policies that would penalize—or even criminalize—bullying were introduced, bullying remains a significant form of in-school victimization among adolescents, with some evidence that the negative effects may be even more exaggerated among LGBTQ students or students of higher body weight. However, due to data limitations, there is limited evidence on the effect of bullying on youth outcomes. It is even harder to assess whether youth across the sexual orientation and weight distribution face different probabilities of being bullied and, thus, different levels of participating in risky activities. In this paper, I allow for unobserved differences between bullied and non-bullied students using mild non-parametric, structural assumptions to estimate bounds within which bullying and cyberbullying victimization may affect participation in a multitude of risk-related activities such as binge drinking, marijuana use, and suicide attempts. Preliminary findings indicate that bullying increases participation in all risk-related activities and that if a student has been at least cyberbullied, is much more likely to engage in such activities. Imposing a further non-parametric, structural assumption—that students will systematically bully other students based on their sexual orientation and weight—I find that this assumption further narrows these bounds and strengthens the above finding: gay students are more likely to commit suicide if they are cyberbullied or are both bullied and cyberbullied, but not if they only experience face-to-face bullying. However, for students who self-identify as obese, any type of bullying increases the probability of participating in a risky behavior. These results suggest that not only more focus should be placed on educating adolescents about the deleterious effects of (cyber)bullying, but also more resources should be allocated towards fighting sexual orientation and weight discrimination among students.
“Do after-school programs affect academic performance? Evidence from the 21st Century Community Learning Centers program”
This paper investigates the impact of before- and after-school academic enrichment programs funded by the 21st Century Community Learning Center (21CCLC) grant on student achievement. The 21CCLC program aims at students in high-poverty, low-performing schools. I analyze the first two cohorts awarded the program in school years 2002/03 and 2003/04 and compare them to public schools not funded by the program. Using difference-in-differences estimation, I find that schools receiving the program experienced a higher percentage of students meeting or exceeding test standards: 1.3% higher in the first year and 2.1% in the second year compared to schools without the intervention. Differentials were highest for middle schools, with the outcome variable measuring 9.0% higher for schools with the intervention. Moreover, difference-in-difference-in-difference analysis at the district level reveals that districts with a higher concentration of minorities and low-income students realized the most benefits from this policy intervention. These results give suggestive evidence about the efficacy of academic enrichment programs, particularly those targeting low-income students.
“Show me your friend to tell you how you fare: peer effects on bullying and academic achievement”
Previous research has focused on whether peers of different backgrounds and abilities directly affect student performance. Contrary to these studies, in this paper, I investigate whether physical school victimization (bullying) affects student academic performance, emphasizing the role of social interactions. Experiences of bullying lead to a disinvestment in human capital, such as performance in standardized math and reading tests. At the same time, some adolescents might respond to increased school victimization based on their peers’ characteristics, which determine the probability of being bullied. That is, peers may negate the negative effects of bullying (higher quality peers) or exaggerate its negative impacts (lower quality peers) on student performance. Because adolescent social networks come primarily from their schools, to identify these effects, I use information on public high school students in eighth and tenth grade from the National Longitudinal Study of Adolescent Health, which includes details on self-identified friends. I start with a linear-in-means specification and an alternative specification with school-specific trends to estimate if peers—and in particular peers who have experienced in-school bullying—affect individual bullying experiences and educational outcomes, including school attendance and GPA. Due to potential endogenous selection of students in schools based on their peers and school safety, I control for school fixed-effects to alleviate this selection bias. Consistent with previous studies on peer effects, preliminary results show that the estimated peer effects are small to non-existent. However, such linear-in-means models do not allow separating peer effects from other confounding factors to identify endogenous and contextual effects separately. To identify these effects, I use spatial autoregressive models with network fixed effects. These nonlinear models resolve the reflection problem and show that there are significant peer (endogenous) and contextual effects on academic performance after controlling for friend networks.
“Do labor market intermediaries affect the speed of employer learning?” (with German Blanco Lobo and Afrin Islam)
In this paper, we examine how employer learning about worker productivity can be affected by labor market intermediaries. The employer learning model suggests that, at the time of hiring, an employer’s ability to screen a worker’s productivity is based on easily observable characteristics (e.g., education level). In a regression framework, after the employer starts learning about the worker’s true productivity, the importance of the easily observable characteristics diminishes, while the importance of hard-to-observe characteristics related to productivity (e.g., ability) rises. Labor market intermediaries have the potential to affect the matching process between workers and employers in various ways. Here, we argue that the extra information possessed by the intermediary could potentially impact both the level at which employer learning about worker productivity starts and the actual speed of the learning process. The intermediary we analyze is job placement services. Information on worker characteristics and usage of placement services comes from the 1979 National Longitudinal Survey of Youth. Since the decision to take advantage of placement services is endogenous, our estimates of employer learning control for sample selection using a control function approach. For the sample of workers using placement services, our results suggest that within three years after hiring, employers learn and correct about half of the initial error in their prediction of worker productivity. This result is in contrast with the four years required for employers that hired workers that did not use placement services.
“Intergenerational transmission of parental care”
There is a significant portion of households—especially in Europe—comprised of extended families, which might suggest that individuals are supporting their parents not only through spending time with them but also by covering their parents’ expenses as part of the household budget constraint. Even for nuclear families, individuals can still support their parents—even though they do not reside together—through financial transfers. Such differences in how individuals support their parents—within household reallocation of resources or across household monetary transfers—may reflect cultural and societal preferences. Especially individuals who are parents themselves need to decide how to allocate their limited time and financial resources towards their own children and supporting their own parents. In this paper, I investigate whether individuals who provide financial support to their parents substitute away from investments in their own children and whether these individuals are also more likely to be financially supported by their own children in the future. To investigate this intergenerational transmission of parental support, I use information on a sample of matched grandparents-parents-children from the German Socio-Economic Panel. I find that individuals whose parents support their grandparents by sending money are more likely to send financial support to their parents and send more money conditional on choosing to support their parents. These results remain even after correcting for potential endogeneity of the decision to provide financial support to their parents using an instrumental variable approach and are robust to estimating propensity score matching models. Moreover, despite evidence of strong (in magnitude) substitutability between sending money to parents and spending time with their own children, the estimated effects are not statistically significant. Further investigation of underlying mechanisms shows that culture and social environment are important in the transmission process, with implications for modeling the socialization production function and for policies that simultaneously address ways to support children and older adults.
“Intergenerational transmission of trust attitudes across four cultures”
Prior research has shown evidence of transmission of skills from parents to their children. More recently, researchers have attempted to understand whether preferences and attitudes (e.g., risk, trust) may also be transmitted to children. Though such studies investigate the relationship between trustworthiness of the parents and their children, they are limited to cross-sectional data, and there is still uncertainty about which mechanisms may lead to children developing different levels of trust. This is particularly important given that a country’s general level of trust may affect economic growth. The goal of this paper is two-fold. I begin with creating mother-child pairs and mother-father-child triads to evaluate whether parental trust leads to child trust and whether this transmission varies with the gender of the parent. I use information from four nationally representative datasets—Swiss Household Panel (Switzerland), German Socio-Economic Panel (Germany), British Household Panel Study (United Kingdom), and 1979 National Longitudinal Survey of Youth (United States)—to examine whether culture affects the degree and the speed of this transmission. Using various methods, including instrumental variables with fixed effects with religion of the parents as an instrument for parental trust, I show that parental trust leads to higher child trust. This transmission is stronger within mother-child rather than father-child pairs, and the effects are stronger at the lower end of the trust distribution. Using lagged values of maternal trust as an alternative source of variation shows that trust transmission starts as early as during early childhood, and there are persistent effects. In the second part of the analysis, I examine if country-wide trust preferences affect the formation and transmission of child trust. To account for such societal trust attitudes, I aggregate the individual-level data at the country level and merge information from the World Values Survey. This two-stage mixed estimation method allows us to examine socialization as a potential transmission mechanism and whether this transmission varies with the level of economic growth.
“Personality traits, occupational choice, and the gender wage gap”
This paper analyses the effect of noncognitive skills on gender wage differences when they affect worker productivity (direct effect) and occupational choice (indirect effect). Using data from the NCDS and jointly modeling gender-specific occupational attainment and wage determination, I find that the magnitude of the noncognitive skills' contribution to the gender wage gap is underestimated by up to 18 percentage points when the indirect effect is overlooked. This contribution differs with age. At age 33, women directly benefit because of higher productivity in noncognitive skills. At age 50, women benefit indirectly because of sorting into occupations that reward these skills.
“Internet health information, consumer types, and health production” (with Molla Mursaleen Shiraj).
In 2012, as many as 35 percent of U.S. adults used the internet to look for health-related information. Because the information provided online can affect an individual’s decision to utilize alternative sources of health information (i.e., physicians), we derive an input demand function for online health information as a function of own costs, costs of complement/substitute sources of information, and health goods. Therefore, our main goal is to identify the types of individuals who seek health-related information through the internet. To answer this question, we use data from the Health Information National Trends Survey (HINTS), which allows us to control for a set of demographic characteristics—capturing differences in the efficiency of the health production process—as well as to evaluate how other sources of information may prevent (or encourage) the use of the internet. However, the choice to utilize online health information may vary based on the severity of the health condition; by using information on common health problems such as headaches or backaches and on more severe health problems such as cancer, we show that there is additional heterogeneity in the types of individuals who choose to use the internet as their primary source of information. Moreover, given that in our model health information is an input in a health production function, we examine whether the source of health information leads to different health behaviors. By comparing individuals who used the internet relative to an alternative source of information, we analyze the impact of the internet on health behaviors such as smoking, nutrition, physical activity, and self-assessed health status. Finally, as a robustness check, we also examine the characteristics of a different cohort of individuals from the Panel Study of Income Dynamics (PSID) and of a different geographic region from the European Social Survey (ESS) to examine whether our results are sensitive to the choice of the data source.
I examine the consequences of enhancing the penalties for people who are HIV-positive and expose others to the infection. Using state-year variation in HIV criminalization laws, difference-in-differences models reveal decreases in diagnoses of new HIV infections and in prevalence of HIV. There is no evidence that enhancements in HIV criminalization laws lead to changes in new AIDS cases, further supported by event study analyses that examine the possibility of delayed policy effects. Exploring within-law variation in the type of penalty (misdemeanor, offense, felony) does not reveal heterogeneous effects, nor does examining heterogeneous effects for sexual minorities and non-sexual minorities. Given these significant effects only for HIV, I use individual-level information to explore changes in HIV testing behaviors and in use of HIV-prevention services as potential mechanisms. The results support that the laws resulted in more frequent testing and use of preventive sexual health medical care and a decrease in the number of sexual partners.
“Anti-bullying legislation and academic performance across OECD countries”
Despite the limited number of studies evaluating the impact of bullying behaviors on schooling, there are no studies on whether such policy interventions directly affect schooling, including performance in standardized math and reading tests and high school and college graduation rates. In addition, no studies evaluate the impact of these policies on multiple countries. I start with difference-in-differences models to show that anti-bullying laws decrease the share of bullying victimization and bullying perpetration across countries. However, there exists heterogeneity based on social attitudes towards victimization. The rest of the paper evaluates whether anti-bullying laws directly affect schooling for these two groups of countries using the same difference-in-differences framework. Event study analyses provide evidence that the laws have an effect on improving educational outcomes, though these effects are stronger for countries where other victimization-prevention policies are lacking.
“The bullying-education gradient: Victims and perpetrators of bullying behaviors in developed and developing countries”
Prior research has uncovered an adverse effect of bullying on academic performance in a few developed countries. However, the evidence is still elusive for most developed countries and non-existent for developing countries. The goal of the paper is twofold. First, I examine what determines the probability of being the victim and the perpetrator of bullying across these two groups of countries. However, because students in international databases are not followed across different years and because they do not include detailed information on academic performance, I aggregate the individual-level data to create (and track) groups of students based on their gender, cohort, and country of residence. In the second part of the analysis, I focus on the direct relationship between bullying rates and schooling. I use a regression discontinuity design to examine whether country-level bullying laws significantly affect bullying rates and whether such laws have differential effects based on the level of economic growth. Then, using compulsory bullying laws as instruments for bullying, I follow synthetic cohorts from different countries to estimate the impact of bullying rates on educational attainment. This paper will allow us to evaluate whether bullying prevalence is detrimental for certain regions across the world and whether economic growth is hindered by bullying victimization by limiting the accumulation of human capital.
“The effect of marriage equality on bullying victimization of sexual minority youth”
Legal inequalities may increase hate crimes because they provide opportunities for discrimination and violence. Although several states introduced state policies recognizing same-sex marriage before 2015, there are still hate crimes against this population group. Among youth, hate crimes may be realized in the form of in-school victimization, with some evidence that the adverse effects may be even more exaggerated among LGBTQ students. However, due to data limitations, there is limited evidence on the effect of bullying on youth outcomes, and it is even harder to assess whether youth of different sexual orientations face different probabilities of being bullied. In this paper, I examine whether public policies toward marriage equality for gay and lesbian rights affect the incidence of bullying victimization based on sexual orientation. Using panel data from 2009 to 2019, I analyze whether state-specific policies for legalization of same-sex marriage before the 2015 federal recognition affect such victimization. I also investigate the degree of tolerance towards LGBTQ individuals—as captured by the amount of other recognized LGBTQ rights (e.g., right to adoption, anti-discrimination laws)—that may affect the incidence of bullying victimization.
“Gay, obese, and bullied? Heterogeneous effects of bullying and cyberbullying on health behaviors”
Even though in the 2000s various policies that would penalize—or even criminalize—bullying were introduced, bullying remains a significant form of in-school victimization among adolescents, with some evidence that the negative effects may be even more exaggerated among LGBTQ students or students of higher body weight. However, due to data limitations, there is limited evidence on the effect of bullying on youth outcomes. It is even harder to assess whether youth across the sexual orientation and weight distribution face different probabilities of being bullied and, thus, different levels of participating in risky activities. In this paper, I allow for unobserved differences between bullied and non-bullied students using mild non-parametric, structural assumptions to estimate bounds within which bullying and cyberbullying victimization may affect participation in a multitude of risk-related activities such as binge drinking, marijuana use, and suicide attempts. Preliminary findings indicate that bullying increases participation in all risk-related activities and that if a student has been at least cyberbullied, is much more likely to engage in such activities. Imposing a further non-parametric, structural assumption—that students will systematically bully other students based on their sexual orientation and weight—I find that this assumption further narrows these bounds and strengthens the above finding: gay students are more likely to commit suicide if they are cyberbullied or are both bullied and cyberbullied, but not if they only experience face-to-face bullying. However, for students who self-identify as obese, any type of bullying increases the probability of participating in a risky behavior. These results suggest that not only more focus should be placed on educating adolescents about the deleterious effects of (cyber)bullying, but also more resources should be allocated towards fighting sexual orientation and weight discrimination among students.
“Do after-school programs affect academic performance? Evidence from the 21st Century Community Learning Centers program”
This paper investigates the impact of before- and after-school academic enrichment programs funded by the 21st Century Community Learning Center (21CCLC) grant on student achievement. The 21CCLC program aims at students in high-poverty, low-performing schools. I analyze the first two cohorts awarded the program in school years 2002/03 and 2003/04 and compare them to public schools not funded by the program. Using difference-in-differences estimation, I find that schools receiving the program experienced a higher percentage of students meeting or exceeding test standards: 1.3% higher in the first year and 2.1% in the second year compared to schools without the intervention. Differentials were highest for middle schools, with the outcome variable measuring 9.0% higher for schools with the intervention. Moreover, difference-in-difference-in-difference analysis at the district level reveals that districts with a higher concentration of minorities and low-income students realized the most benefits from this policy intervention. These results give suggestive evidence about the efficacy of academic enrichment programs, particularly those targeting low-income students.
“Show me your friend to tell you how you fare: peer effects on bullying and academic achievement”
Previous research has focused on whether peers of different backgrounds and abilities directly affect student performance. Contrary to these studies, in this paper, I investigate whether physical school victimization (bullying) affects student academic performance, emphasizing the role of social interactions. Experiences of bullying lead to a disinvestment in human capital, such as performance in standardized math and reading tests. At the same time, some adolescents might respond to increased school victimization based on their peers’ characteristics, which determine the probability of being bullied. That is, peers may negate the negative effects of bullying (higher quality peers) or exaggerate its negative impacts (lower quality peers) on student performance. Because adolescent social networks come primarily from their schools, to identify these effects, I use information on public high school students in eighth and tenth grade from the National Longitudinal Study of Adolescent Health, which includes details on self-identified friends. I start with a linear-in-means specification and an alternative specification with school-specific trends to estimate if peers—and in particular peers who have experienced in-school bullying—affect individual bullying experiences and educational outcomes, including school attendance and GPA. Due to potential endogenous selection of students in schools based on their peers and school safety, I control for school fixed-effects to alleviate this selection bias. Consistent with previous studies on peer effects, preliminary results show that the estimated peer effects are small to non-existent. However, such linear-in-means models do not allow separating peer effects from other confounding factors to identify endogenous and contextual effects separately. To identify these effects, I use spatial autoregressive models with network fixed effects. These nonlinear models resolve the reflection problem and show that there are significant peer (endogenous) and contextual effects on academic performance after controlling for friend networks.
“Do labor market intermediaries affect the speed of employer learning?” (with German Blanco Lobo and Afrin Islam)
In this paper, we examine how employer learning about worker productivity can be affected by labor market intermediaries. The employer learning model suggests that, at the time of hiring, an employer’s ability to screen a worker’s productivity is based on easily observable characteristics (e.g., education level). In a regression framework, after the employer starts learning about the worker’s true productivity, the importance of the easily observable characteristics diminishes, while the importance of hard-to-observe characteristics related to productivity (e.g., ability) rises. Labor market intermediaries have the potential to affect the matching process between workers and employers in various ways. Here, we argue that the extra information possessed by the intermediary could potentially impact both the level at which employer learning about worker productivity starts and the actual speed of the learning process. The intermediary we analyze is job placement services. Information on worker characteristics and usage of placement services comes from the 1979 National Longitudinal Survey of Youth. Since the decision to take advantage of placement services is endogenous, our estimates of employer learning control for sample selection using a control function approach. For the sample of workers using placement services, our results suggest that within three years after hiring, employers learn and correct about half of the initial error in their prediction of worker productivity. This result is in contrast with the four years required for employers that hired workers that did not use placement services.
“Intergenerational transmission of parental care”
There is a significant portion of households—especially in Europe—comprised of extended families, which might suggest that individuals are supporting their parents not only through spending time with them but also by covering their parents’ expenses as part of the household budget constraint. Even for nuclear families, individuals can still support their parents—even though they do not reside together—through financial transfers. Such differences in how individuals support their parents—within household reallocation of resources or across household monetary transfers—may reflect cultural and societal preferences. Especially individuals who are parents themselves need to decide how to allocate their limited time and financial resources towards their own children and supporting their own parents. In this paper, I investigate whether individuals who provide financial support to their parents substitute away from investments in their own children and whether these individuals are also more likely to be financially supported by their own children in the future. To investigate this intergenerational transmission of parental support, I use information on a sample of matched grandparents-parents-children from the German Socio-Economic Panel. I find that individuals whose parents support their grandparents by sending money are more likely to send financial support to their parents and send more money conditional on choosing to support their parents. These results remain even after correcting for potential endogeneity of the decision to provide financial support to their parents using an instrumental variable approach and are robust to estimating propensity score matching models. Moreover, despite evidence of strong (in magnitude) substitutability between sending money to parents and spending time with their own children, the estimated effects are not statistically significant. Further investigation of underlying mechanisms shows that culture and social environment are important in the transmission process, with implications for modeling the socialization production function and for policies that simultaneously address ways to support children and older adults.
“Intergenerational transmission of trust attitudes across four cultures”
Prior research has shown evidence of transmission of skills from parents to their children. More recently, researchers have attempted to understand whether preferences and attitudes (e.g., risk, trust) may also be transmitted to children. Though such studies investigate the relationship between trustworthiness of the parents and their children, they are limited to cross-sectional data, and there is still uncertainty about which mechanisms may lead to children developing different levels of trust. This is particularly important given that a country’s general level of trust may affect economic growth. The goal of this paper is two-fold. I begin with creating mother-child pairs and mother-father-child triads to evaluate whether parental trust leads to child trust and whether this transmission varies with the gender of the parent. I use information from four nationally representative datasets—Swiss Household Panel (Switzerland), German Socio-Economic Panel (Germany), British Household Panel Study (United Kingdom), and 1979 National Longitudinal Survey of Youth (United States)—to examine whether culture affects the degree and the speed of this transmission. Using various methods, including instrumental variables with fixed effects with religion of the parents as an instrument for parental trust, I show that parental trust leads to higher child trust. This transmission is stronger within mother-child rather than father-child pairs, and the effects are stronger at the lower end of the trust distribution. Using lagged values of maternal trust as an alternative source of variation shows that trust transmission starts as early as during early childhood, and there are persistent effects. In the second part of the analysis, I examine if country-wide trust preferences affect the formation and transmission of child trust. To account for such societal trust attitudes, I aggregate the individual-level data at the country level and merge information from the World Values Survey. This two-stage mixed estimation method allows us to examine socialization as a potential transmission mechanism and whether this transmission varies with the level of economic growth.
“Personality traits, occupational choice, and the gender wage gap”
This paper analyses the effect of noncognitive skills on gender wage differences when they affect worker productivity (direct effect) and occupational choice (indirect effect). Using data from the NCDS and jointly modeling gender-specific occupational attainment and wage determination, I find that the magnitude of the noncognitive skills' contribution to the gender wage gap is underestimated by up to 18 percentage points when the indirect effect is overlooked. This contribution differs with age. At age 33, women directly benefit because of higher productivity in noncognitive skills. At age 50, women benefit indirectly because of sorting into occupations that reward these skills.
“Internet health information, consumer types, and health production” (with Molla Mursaleen Shiraj).
In 2012, as many as 35 percent of U.S. adults used the internet to look for health-related information. Because the information provided online can affect an individual’s decision to utilize alternative sources of health information (i.e., physicians), we derive an input demand function for online health information as a function of own costs, costs of complement/substitute sources of information, and health goods. Therefore, our main goal is to identify the types of individuals who seek health-related information through the internet. To answer this question, we use data from the Health Information National Trends Survey (HINTS), which allows us to control for a set of demographic characteristics—capturing differences in the efficiency of the health production process—as well as to evaluate how other sources of information may prevent (or encourage) the use of the internet. However, the choice to utilize online health information may vary based on the severity of the health condition; by using information on common health problems such as headaches or backaches and on more severe health problems such as cancer, we show that there is additional heterogeneity in the types of individuals who choose to use the internet as their primary source of information. Moreover, given that in our model health information is an input in a health production function, we examine whether the source of health information leads to different health behaviors. By comparing individuals who used the internet relative to an alternative source of information, we analyze the impact of the internet on health behaviors such as smoking, nutrition, physical activity, and self-assessed health status. Finally, as a robustness check, we also examine the characteristics of a different cohort of individuals from the Panel Study of Income Dynamics (PSID) and of a different geographic region from the European Social Survey (ESS) to examine whether our results are sensitive to the choice of the data source.