Controls for individual characteristics include an indicator of whether the student separated from their school in an earlier school year, and indicators for gender, Aboriginal identity, language spoken at home (any language besides English), disability, giftedness, ESL status, and the student’s own FSA exam score. We also control for a set of socioeconomic characteristics of the Census Enumeration or Dissemination Area (EA or DA, respectively) in which the student resides as proxies for unobserved student background characteristics. Specifically, we control for the proportion of household heads in the EA/DA who immigrated to Canada in the previous five years; whose education level was less than grade 9, without a high school diploma, with a high school diploma, and with a bachelor’s degree or higher (the omitted category is those with more than high school but less than a bachelor’s degree); who are visible minority; who are single parents; who moved into the EA or DA in the last year or in the last five years; and the unemployment rate among those over age 25, the average dwelling value, average household income, and the fraction who own their dwelling. Details of the construction of these variables are provided below.
The present work is related to two different literatures: studies on women’s empowerment and studies on early childhood development. The literature on women’s empowerment is more extensive. Kabeer (2005) defines empowerment as the “ability to make choices” in ways that change power relations and affect women’s education, employment, and politi- cal participation. Duflo (2012) defines women’s empowerment as improving women access and utilization of social, political and economic opportunities. Decision-making within the household is also an important indicator of the distribution of power within the household (Alkire, 2007; Narayan-Parker, 2005). We will follow this approach and explore intra- household decisions, capturing women’s power relations within the household and their access to the constituents of development (e.g. whether they are allowed to work). Differ- ent channels for empowering women have been explored. Education is sometimes proposed as one of the main drivers of empowerment (Oyitso and Olomukoro, 2012) , but the evi- dence is mixed. There is substantial evidence that education can improve cognitive skills, raise aspirations, allow access to information, raise awareness to real conditions, and help coping with dis-equilibrium (Kabeer, 2005; LeVine et al., 2001). More educated women also experience less domestic violence (Kabeer, 2005; Sen, 1999). Mocan and Cannonier (2012) find that more educated women in Sierra Leone are “more intolerant of practices that conflict with their well-being”. It is less clear, however, whether this change in prefer- ences translates into behavior. Andrabi et al. (2012) show that higher maternal education improves maternal child care, but the study does not find an effect on intra-household decision-making.
Columns 1 through 3 show that peers’ BMI does not seem to influence student’s daily meals. The results may be due to the responses themselves being not quite relevant; admit- tedly, eating dinner together with or without family could not affect student’s BMI much. Nevertheless, Columns 1 and 2 indicate that peers would not affect student’s number of having meals or appetite. Columns 4 through 6 show that peers’ BMI does not affect the student’s time use in indoor or outdoor activities. The dependent variable in Column 6 is the sum of minutes spent per day on doing homework, reading books, and self-study. Though we regress time use in those three study related activities separately, we do not find any significant effects. These results are in line with Yakusheva, Kapinos and Eisen- berg (2014)’s finding that their peer effect on weight gain is not driven by peer’s exercise habits or eating disorder symptoms.
including interactions between the teacher experience variable and the proportions of disabled peers. The interaction between undiagnosed peers and the proportion of novice teachers is large and negative for math and English test scores and positive for suspensions, although it is only statistically significant for math. Nevertheless, these results provide suggestive evidence that more experienced teachers may be better at mitigating negative peer effects than inexperienced teachers, possibly because they are better able to minimize disruptions. Non-disabled students may also benefit if disabled students are educated in separate classrooms, as are the majority of special education students in New York City. Even excluding the special education only schools in District 75, about 78% of special education students attend special education only classes in math or reading. While students who enter or exit special education are more likely to be mainstreamed, only 32% of them attend general education classes when they receive special education services. I take two approaches to determine whether the segregation explains why students with undiagnosed disabilities have a negative effect on their peers, but special education students do not. First, I use classroom identifiers to determine the following proportions of school-grade peers: (1) the proportion who have undiagnosed disabilities and will be placed in special education only classes in the future, (2) the proportion currently attending special education only classes, and (3) the proportion who have been declassified but attended special education only classes in the past. In Table 2.5, I use the same subsample of students as Table 2.4 and include (1) through (3) as additional regressors in the main specifications (Columns 1, 3, and 5). The coefficients on (1) through (3) test whether disabled students who are ever segregated have different effects than those who are not. 18 The estimates suggest that disabled students
nomic growth (Moretti, 2004). Therefore, in the second essay, I investigate how college education affects people’s cross-province migration. Since the expansion of higher ed- ucation resulted in a large difference in the openings of colleges and universities across provinces, I am able to use the expansion as exogenous variation in educational attain- ment to identify the impact of college on migration. Consistent with what is found in the U.S. and Europe (Malamud and Wozniak, 2012; Machin, Pelkonen, and Savanes, 2012; B ¨ ockerman and Haapanen, 2013; Weiss, 2015), college education increases Chinese young adult’s out-province migration rate by 9.1 percentage points. Considering the average mi- gration rate is around 9 percent, this effect doubles the migration propensity for college educated people. One limitation of this essay is that it could I could not identify where the individual attended college. Hence, it is not clear whether some of the migration effect occurs because individuals relocate to attend college or is concentrated on relocation after college. I leave this question for future research. In order to examine the mechanism that facilitates college educated people move more, I utilize several survey questions related to attitudes on success in life, the important of education, and the role of family network. Instrumental variables estimates show that people with college experience rely less on family and are more prone to agree that education is more important. One unique fea- ture of the Chinese institutional setting is the hukou system. Because moving in China is no longer restricted by hukou status, I do not focus on migration measures related to hukou registered location. However, I do find that people who held rural hukou sta- tus during childhood are more likely to conduct long-distance movement after attending college compared to their urban counterparts. The underlying policy implication is infor- mative. The opportunity cost associated with changing hukou status is big. Because rural hukou holders face smaller cost of hukou status regardless of which urban city they go, they may have higher mobility than urban people who have been enjoying local hukou benefits. Further reforming hukou policies to motivate more educated urban people to move across geographical areas is important to spatial allocation of skills.
The attendance data are augmented by student demographic and achieve- ment data. Demographic data comes from LUSD’s student entrance file. The entrance file includes all students who enter into a traditional public high school between the years 2005 and 2009. Dropout and transfer outcomes are from the student exit file, which contains detailed administrative exit information gener- ally unavailable in survey datasets. It distinguishes between types of transfers (in distinct, public in state, private in state, out of state, out of country) and types of diplomas that students earned. Diploma types include a high school diploma from a traditional school, a high school diploma from an alternative school, an adult education diploma, a GED, or no diploma. Alternative schools are high schools within LUSD that waive some of the requirements of traditional high schools. They can have non-standard school days, online learning components, lower credit requirements for graduation, and may waive some higher level classes, especially in math, that are otherwise required for graduation. Alter- native schools do not include other schools of choice within the district such as magnet, gifted, or area specialty schools which have the same requirements as traditional neighborhood public schools. Diplomas from alternative schools are likely to have lower value than those from traditional public high schools due to their reduced requirements. Furthermore, nearly half of students in the sample who transfer to an alternative high school eventually drop out.
These statistics and facts show that the unemployment levels in European countries have behaved differently over the years and this can lead to greater divergences in labour market across Europe as a whole. The role played by flows into and out of employment is one of the possible determinants behind the variations in European unemployment rate. Labour markets are in a constant state of flux, and the same stocks and structures at the aggregate level can be linked to diverse patterns of employment transitions. I analyse annual worker flows and the composition of these flows for different worker groups over Europe and across various European countries for over a period of 2006-2016. My analysis answers the following two questions. The first question is whether high unemployment in Europe is a result of high job separation or low job finding. The second question is how important are socio-demographics, immigration status and degree of urbanisation in shaping worker flow patterns in Europe? Using the yearly data from European Union Labour Force Survey (EU-LFS), I compute European worker flows to better understand the differences in unemployment rates over time and most importantly to what extent does the change in aggregate unemployment rates are attributed to each of these flows. I then evaluate the correlation between various socio- demographic characteristics (SDCs), immigration status and degree of urbanisation on employment transitions to demonstrate that in addition to basic economic factors that affect the labour demand, changing demographic characteristics of workforce may also influence on European unemployment.
Moreover, parental education plays a significant role in educational success in Hungary. For instance, the incremental (mean) Reading score for children whose parents (at least one parent) hold (s) a high school degree relative to those whose parents at most finished primary school is around 89 percent of the standard deviation of the Reading scores. In addition to other factors, this may be driven by the fact that children from highly educated families are more likely to be engaged in activities that promote academic success. Although the direct impact of such parental input on test scores is difficult to pin down, there are numerous variables in the PIRLS dataset that indicate a positive association between parental education and home activities which promote academic success. For example, whereas approximately 58 percent of the students with parents having at most primary school degree reported that they are often told stories at home, the corresponding figure for students with parents who possess a college or university degree is 83 percent. Among others, Elder and Lubotsky (2006) for the US and Fertig and Kluve (2005) for Germany find evidence for the importance of parental education for schooling success. The latter two authors, based on the “Young Adult Longitudinal Survey” covering 18 – 29 year old individuals, find that in both former East and West Germany, children from low educated families (whose parents at most completed the Hauptschule) are less likely to attain a high school degree (Abitur) and the opposite is true for their counterparts from high educated families (whose parents completed more than Hauptschule). Another piece of evidence which shows the importance of socio- economic background for academic success in Hungary from another perspective merits comment: the comparative analysis of OECD countries implies that the relationship between (15 year old) students’ socio-economic background and their expectations to complete tertiary education is the strongest in Hungary among OECD countries (Education at a Glance 2007 (2007)).
In this paper, I have argued that own-account workers and employers are differ- ent and need to be treated separately. Having a large share of self-employment and small business owners does not mean a country is more entrepreneurial. I have argued that financial intermediation can help explaining the cross coun- try differences in occupation shares. A lower financial intermediation efficiency leads to a higher cost of borrowing and lower return on savings. Agents who save with financial intermediaries are more likely to seek alternative occu- pations to manage more wealth. Wage workers are more likely to become an own-account workers, operating small businesses to manage their wealth. Agents who need to borrow to become employers choose to be own-account worker instead. The result is less capital and labour intermediated through the market. The quantitative results presented here has shown that, by varying financial efficiency, the model can account for over 70% of the cross country variation in the share of own-account workers.
Furthermore, another survey by the NSO called Health and Welfare Survey (HWS) is employed to assess a relationship between education and health outcome in Thai- land. HWS was first conducted annually from 1974 to 1978. Then it was reduced to once in every five years from 1981 to 2001. Since HWS is one of a very few sur- veys records statistics such as number of illness, access to health facilities and type of health insurance, after the implementation of universal health coverage scheme in October 2001, the Ministry of Public Health requested the NSO to conduct HWS an- nually from 2003 to 2007 (National Statistical Office, 2008). Although the NSO has conducted this survey for many years, the questionnaire, especially for proxies of health outcome, changes from one survey to another. Therefore, this preliminary study decide to concentrate on a self-assessed health status by each respondent. Unfortunately, only HWS 2006 contains such a question. Specifically, it asks every respondent of 15 years of age or older what do they think of their health status by rating from 1 (very bad) to 5 (very good). There are 74,057 persons in the survey but the number of people in these cohorts (cohort 1961 - 1970) who respond to the self-assessed health status question are just around 10%. The descriptive statistics of the sample (by gender) are presented in tables 3.10 - 3.12.
Note: This table presents regression analysis between trust and admiration for a longer time period. The data is obtained from the General Social Survey, Annual National Electoral Survey and Gallup Opinion Polls. The data spans from 1972-2012, 1964-2008 and 1948-2013 respectively. All regressions control for region and year fixed effects. Robust standard errors applied for ANES sample. Standard errors clustered at the state level for the rest of the regressions. Linear probability models estimated for columns (1) to (4), while multinomial logit specification estimated for columns (5) and (6) with referent category as Don’t Know. Non White is a dummy variable for race, Age is measured in years, Male is a dummy variable for gender, Catholic, Jewish and Other and None are dummy variables for religious affiliation (relative to Protestants), Democrat and Other Party are dummy variables for respondents political preferences (relative to being Republican), Democrat/ Other Party X Republican is an interaction between respondent’s political preference and if the sitting President is a Republican and college is a dummy variable for having some years of post-secondary education. The beta proportionality test reports the chi-square statistic. Trust in Federal Government in the GSS sample is a measure of confidence in the Executive branch of Federal Government. Beta correlations are correlations between the predicted values across the regressions.
Other economic variables are taken as control variables. According to the theory of expected utility, a wealthy individual is more likely to be less risk averse. Additionally, accessing high risk assets is more costly either because of asset prices or entry fees, implying that the range of options of financial risks is also subject to the investor’s disposable resources. We choose two continuous variables, (1) labor income and net worth in the last calendar year, a binary variable, and (2) credit constraint, as proxies of disposable resources. Labor income includes wage, professional income from a sole proprietorship or a farm, compensation and income from social security, pensions and kindred programs. Net worth comprises two parts: financial and nonfinancial net worth. Financial net worth is the sum of the current values of financial assets minus loans still owed for purchasing financial assets. Nonfinancial net worth is the sum of net principal residence, net nonprincipal residence, net other properties, and net business capital minus the sum of credit card loans, education loans and any other loans. 44 The definition of credit constraints is imputed from the SCF questionnaire. Households who have been denied loans from financial institutions or do not borrow because they expect to be denied are defined to be credit constrained.
To identify the effect of FSP participation on the probability that a mother becomes obese I use data from the 1986-2004 waves of the NLSY79 to construct two samples—the first is a sample of all mothers and the second a sample of low-income mothers. A woman enters the sample the first time she reports living with a child, regardless of their biological relationship, and remains in the sample until failure or censoring occurs. In 1986 there were 6,298 women, of whom 1,377 were eliminated because they had missing or invalid BMI observations. Next women whose education level is not observable were eliminated from the sample, leaving 4,396 women. In 1986 the average age of women in this sample was 25.6 years and 50.2% of women were already mothers. Nearly 40% of women were participating in the FSP, and of those participating in the FSP nearly 64.0% were mothers. On average, these women had completed 12.8 grades of school. Summary statistics for the full sample are presented in the first column of Table 3.2.
Given the critical role of teacher quality on pupil’s performance and evidence on the correlation between teachers’ education and learning, recruiting and maintaining the most efficient teachers should be prioritised. The issue is how to attract and select good teachers. This is not a straightforward process since it is difficult to assess ex-ante if a candidate would be a good teacher. Qualitative research suggests that top-performing school systems manage to attract better people into the teaching profession, leading to better student outcomes. They do this by introducing highly selective teachers training, developing effective selection processes for identifying the right candidates and paying good (but not great) starting compensation. Conversely, lower-performing school systems rarely attract the right people into teaching. The success in attracting talented people into teaching is linked to specific country features such as history, culture and status of teaching profession. However, there are some policies that can be implemented to attract the best graduates, such as effective mechanisms for selecting teachers, good teacher training programmes, good starting compensations and increasing professional autonomy in
This dissertation is comprised of threeessays, each studying a different aspect related to topics in the economics of education. These pieces each study how education production is impacted by different economic forces. In Chapter 1, I study the effect of racial segregation on academic achievement, college preparation, and postsecondary attainment in a large, urban school district. To achieve racial balance in its oversubscribed magnet schools, this district conducted separate admissions lotteries for black and non-black students. Because the student body was predominantly black, administrators set aside disproportionately more seats for the non-black lottery. In 2003, the federal Office of Civil Rights forced this district to instead use a race-blind lottery procedure that dramatically increased racial segregation for incoming magnet school cohorts. In an instrumental variables framework that exploits both randomized lottery offers and this unanticipated shock to racial makeup, I test whether student racial composition is a meaningful input in the education production function. As a baseline, I use admissions lotteries to estimate the effect of enrolling in a magnet middle school on student outcomes. In general, enrollment returns are comparable between magnet and traditional schools, but I estimate heterogeneous magnet school effects across student subgroups. Education production is sensitive to school racial composition in that segregation has a deleterious impact on student outcomes. I find that increasing the share of black peers in a cohort decreases student achievement in math, science, and writing for black students with losses primarily driven by high-aptitude black students. Further, racial segregation erodes high school graduation rates and also decreases college attendance by reducing enrollment at 2-year institutions among female black students. These findings suggest that policies aimed at achieving racial balance in schools will likely increase aggregate educational achievement.
amount of funds to induce more graduates in STEM (Atkinson and Mayo, 2010). The US federal government for instance is considering actions with the objective of increasing STEM graduates by 34% annually (The President’s Council of Advisor on Science and Technology, 2012). Still, the graduation rate or even the degree of interest for students in graduating in these majors has remained pretty stable since the ’80s (Altonji et al., 2012) and, while the literature on choices of educational levels is very wide and consolidated (starting from the seminal work by Mincer (1974)), there is relatively little work on choices of fields of study. This paper evaluates how much of the lack in STEM graduates can be attributed to high schools, and in particular to the curriculum they offer. Ellison and Swanson (2012) show indeed that there is a large heterogeneity in high schools effectiveness in developing talents in math and science, which is not explained by differences in schools composition. This paper investigates the role of high school curriculum and it addresses three questions in particular. First, does being exposed to more science courses in high school increase by itself the supply of STEM graduates? Second, who is induced to take more science classes, when exposed to the option of taking them at age 14? Third, this paper evaluates whether more exposure to science at age 14 works for everybody, or whether the effect is concentrated on some segments of the population. This is a relevant question to distinguish whether is it more efficient to force all students to take more science courses in high school or to target the offer
Most educational systems rely on lectures and class meetings as a means of in- struction. This is even more prevalent when secondary or pre-tertiary education is considered. Nevertheless, class attendance is not always perfect. Lecture learning is based on group learning, which may not be the optimal learning style for everyone. As a result, many students decide to skip class when given the opportunity. In a classroom, students compete for the attention and time of the instructor. Thus, their consumption of education induces externalities on one another. Romer  claims that college students in three elite U.S. universities were found to perform better when attending classes and completing homework. Nevertheless, this claim may apply for only a small part in the right tail of the ability distribution in a given society. Lectures in classrooms with samples that reflect the actual ability distri- bution of students may not run completely smoothly. To give an example, students who act up or disrupt the lecture may be more likely to be found in non-elite schools. The question that arises here is whether someone should attend class or stay and study at home given their ability ceteris paribus.
We estimated these interaction models separately for the three school choice programs as well. Results are not shown but are available on request. It was more frequently the case that interactions were significant in these models but the patterns varied across outcomes and types of school choice in ways that suggested at most idiosyncratic effects. The percentage of interactions that were significant at the five percent level were 6.3% for Magnet, 11.5% for Choice and 21.8% for VEEP. These first two percentages are close to the type I error rate expected. VEEP was the outlier, and had two outcomes that accounted for about half of the significant effects. These two grade 12 outcomes — the percentage of days attended and citizenship grades — showed significant positive interactions between winning the lottery and the differences between the choice school and the local school in reading scores, math scores, reading value-added, mean GPA, the proportion of students who were white and the proportion of parents with a Bachelor’s or higher. While all of these relationships are easy to motivate, for example in terms of positive peer effects, these patterns did not show up for other high school or postsecondary outcomes, with a few exceptions.
These exercises demonstrate a strong association between the establishment and intensity of family planning programmes with the decline in fertility rates, after adjusting for changes in per capita income, urbanisation, infant mor- tality, female labour force participation, and educational attainment. Most sub-Saharan African governments acknowledged rapid population growth as a policy concern much later than developing countries elsewhere. Even after the formulation of population control policies, commitment to family planning lagged behind that of other regions leading most international agencies work- ing in family planning to invest their resources in the more promising areas of Asia and Latin America. The onset of the HIV/AIDS epidemic is also likely to have weakened the emphasis on fertility control due to limited resources being targeted towards addressing the epidemic as well as the emergence of a pro-natalist response to the high mortality rates caused by the epidemic (Na- tional Research Council Working Group on Factors Affecting Contraceptive Use 1993). While almost all African countries now provide direct or indirect support for family planning, their efforts have only recently caught up with the rest of the world. Perhaps not surprisingly in light of the strong correlations, the countries in sub-Saharan Africa tend to be the ones where fertility rates still remain above the world’s average.
This possibility requires serious consideration about how to address it. First, we include control variables for which we believe the conditions above will hold. These controls include those considered by Butrica and Karamcheva (2013): parental demographic characteristics (sex, age, if the respondent has reached the age for Social Security uptake, and race and ethnicity), if a spouse is present, reported health status, wealth and income information (the family income excluding individual i’s labor income, and household assets) and state of residence. We differ from Butrica and Karamcheva in three ways: (1) we use the age of Social Security eligibility (62) rather than the Full Retirement Age for Social Security (65) because our sample skews much younger, (2) we do not include data on spousal income or retirement status because these fields are collinear with spouse presence within the PSID data, and (3) we use state rather than Census region because this more granular location data is available in the PSID.