Honors Thesis - Economics
Transcription
Honors Thesis - Economics
Federal Work-Study versus Off-Campus Jobs: Does the type of job matter for post-graduate outcomes? Larissa Muramoto§ Honors Thesis Advisor: Professor Caroline Hoxby June 1, 2013 Abstract The Federal Work-Study (FWS) Program was established as part of the Economic Opportunity Act of 1964 to help students find part-time employment to cover the cost of their tuition. Today, the U.S. government allocates nearly $1 billion in FWS funds for students in 3,000 institutions around the country. Even though the FWS program is large, it has not been studied comprehensively because of difficulties finding exogenous variation that explains endogenous work decisions. To overcome this issue, I exploit differences in FWS availability and FWS availability relative to other job opportunities at comparable institutions to instrument for FWS participation and hours worked in school. I find some evidence that FWS participation improves academic and post-graduation outcomes. Keywords: Work-study, employment, post-graduate outcomes, job quality § Department of Economics, Stanford University, Stanford, CA 94305 (email: larissam@stanford.edu). I would like to thank Caroline Hoxby for her guidance and support. I would also like to thank Michael Dinerstein for his regular feedback throughout this project. Larissa Muramoto 6/1/2013 2 Introduction The Federal Work-Study Program was established as part of the Economic Opportunity Act of 1964 to help students find part-time employment to cover the cost of their tuition. Today, the FWS Program is a significant source of student aid. In 2012, the U.S. government allocated nearly $1 billion in Federal Work-Study funds for students in 3,000 institutions around the country (Federal Campus-Based Programs Data Book 2012). According to the Beginning Postsecondary Students Longitudinal Study, approximately 40% of college students hold a FWS job. In fact, the FWS accounts for 45% of all the jobs college students hold. Even though FWS is such a large program, relatively little has been done to analyze the impact it has on students' achievement during and after college. The potential effects are not obvious: on one hand, the FWS Program might harm students by encouraging them to take up part-time employment that could compete with classes and homework for time. Theoretically, it could be optimal for students to take out larger loans instead of working part-time: with less time to focus on schoolwork, students with part-time jobs might suffer from lower GPAs, take longer to graduate, or even drop out of school. On the other hand, the FWS Program might help students by providing access to convenient jobs on-campus. One could argue that students will work during college regardless of whether FWS jobs are available or not: majority (~70%) of college students hold part-time jobs and more than half of these are not affiliated with the FWS Program (Beginning Postsecondary Students 2012). By providing jobs that are intended to compliment a student’s education, the FWS Program might benefit these students who choose to work. Compared to alternate jobs that require students to commute off campus, FWS jobs can benefit students by providing access to jobs that are more compatible with students' unique college Larissa Muramoto 6/1/2013 3 schedules. Additionally, having a FWS job might increase a student’s persistence via attachment to the institution (Tinto 1987). A student who is integrated into the environment at a college is less likely to drop out when he/she regularly interacts with peers on campus. In the past, it has been difficult to establish a causal relationship between FWS Program participation and outcomes because it is difficult to find exogenous variation that explains endogenous work decisions. A student’s decision to work in college is rarely random: he or she may need the extra finances to fund his or her education. Alternately, he or she might possess higher levels of motivation that cause him or her to seek early work experience to improve job market candidacy. Because it is difficult to identify exogenous variation in work decisions, most existing research on the effects of student employment has simply examined correlations between student employment and academic outcomes. A comprehensive literature review in 2009 by Hossler et al. found the effects of the FWS Program to be inconclusive. Some articles argued that FWS might have positive effects on student outcomes via Tinto’s Theory of Student Departure: participating in the program helps students integrate into the campus community, thereby lowering their chances of dropping out. However, this conclusion was highly subject to question because the authors found it difficult to quantify “integration” into a college environment. Other articles found that work-study had no effect, perhaps because students opt for off-campus jobs that often pay better than work-study jobs and have no cap on hours worked. Hossler et al. concluded that more research needed to be done on the FWS program to assess its true effects on student performance. Economists have also attempted to derive a causal relationship between student employment and outcomes by utilizing instruments such as parental schooling, parental religion, Larissa Muramoto 6/1/2013 4 and net schooling costs (DeSimone 2008; Kalenkoski and Pabilonia 2010). However, these instruments are questionable because it is not obvious that they affect student outcomes solely through their impact on part-time employment in college. For example, it could be argued that parental schooling could impact student outcomes because more educated parents are more likely to educate their children at home (Davis-Kean 2005). The strongest instrument for student employment devised so far is the variation in FWS funds allocated to colleges by Scott-Clayton (2011). This instrument is promising because the attending a college with a higher FWS allocation is likely to encourage students to participate in the FWS Program, given that they are eligible. If more FWS jobs are available on campus, one can expect that more students will participate in the program. Additionally, FWS allocation is also a promising instrument because it is not inherently obvious that the availability of FWS funds is correlated with other student outcomes. Utilizing FWS allocation as an instrument for participation in the FWS Program, Scott-Clayton found that work-study does not have any positive effects on college GPA, credits earned by semester, probability of dropout, or probability of completing a degree in college. Stinebrickner and Stinebrickner (2003) employed an alternate approach to understand the effects of student employment. They exploited data from students at a private college in Kentucky that randomly assigned students to on-campus jobs. Variation in the hours worked stemmed from variation in job assignment: some jobs gave students the choice to work more, while others were limited to 10 hours per week. The result of this study showed that those students who had the opportunity to work more and chose to do so suffered from a decline of 0.162 GPA points per additional work hour. While this result is interesting, it is difficult to draw Larissa Muramoto 6/1/2013 5 larger conclusions because the students who choose to attend such a unique college might not be comparable to college students around the nation. More research is needed to truly understand the causal effects of the FWS Program. While Scott-Clayton and Stinebrickner et al. have provided insight into these effects, both researchers use small datasets that might not be generalizable to the entire US student population. Additionally, there are still no papers examining the difference between FWS jobs and off-campus jobs because of the past difficulties finding an exogenous source of variation in student employment. One still needs to determine whether FWS might have positive benefits after a student graduates because work-study jobs differ from alternate jobs a college student might take such as waitressing or being a cashier. Since work-study jobs are meant to compliment a student’s education, students who participate in work-study are might have advantages over their non-work-study peers in job prospects and preparation for entering the workforce. The Federal Work-Study Program Student Eligibility and Receipt of Funds A student’s eligibility is determined by financial need and enrollment in a Title IV-Eligible postsecondary institution. Financial need is determined by the Free Application for Student Aid (FAFSA), which computes a student's “expected family contribution” (EFC) based on family income, assets, and other variables. The amount of financial aid a student receives is determined by the difference between a student's EFC and the cost of attending an institution. The opportunity to Larissa Muramoto 6/1/2013 6 participate in the FWS Program is often awarded to students as a part of a larger financial aid package (Federal Student Aid Handbook 2003). Unlike Pell Grants or student loans, FWS funds are not directly allocated to students. Rather, FWS funds are allocated to post-secondary institutions that subsequently distribute the funds to their students via government-subsidized part-time employment opportunities on-campus (and occasionally off-campus in conjunction with community service organizations). In return, participating institutions must offer students jobs that are complimentary to college education. Some examples of these jobs are secretarial work for an academic department on campus or tutoring jobs. Since schools determine the distribution of FWS funds, variation in the student receipt of funds arises. For example, some schools may choose to use the funds to provide fewer, but betterpaying jobs for the most needy students while other schools may choose to use the funds to provide more, but lower-paying jobs for students in a wider range of need. Alternately, some schools may give preference to students with higher financial need, better academic scores, year in school, or indicated interest in the FWS on their FAFSA when distributing funds (Scott-Clayton 2011). Since only 16% of schools award FWS to every eligible student, the variation in FWS fund distribution can be significant: it is easy to see why the only reason a student might not receive FWS funding is simply because of the college he or she attends. Institution Eligibility and Receipt of Funds FWS allocation is currently based on a combination of an institution’s original, or “base guarantee” allocation and an institution’s current need. State review boards established the original “base guarantee” allocations in 1965 according to institutions’ grant submissions. Larissa Muramoto 6/1/2013 7 Initially, an institution’s yearly allocation was based on its prior allocation. However, in the 1970s, these review boards were criticized for being inefficient, inconsistent, and easily manipulated by grant writers (Smole 2005). Critics argued that the “base guarantee” allocation was unfair because it gave richer schools an advantage since they had the resources to employ experienced proposal writers that could manipulate the review board. Ironically, the FWS funds were allocated away from smaller, poorer institutions in favor of larger institutions with more resources. As a result, the FWS allocation rule was adjusted to include an “allocation adjustment” that addresses institutional need. Policymakers planned to slowly transition from the “base guarantee” allocation method, but institutional resistance halted the phase-out. Today, approximately 2/3 of FWS funds are distributed based on the “base guarantee” allocation while 1/3 is distributed based on the allocation adjustment. While this is unfortunate for smaller, poorer, and newer schools, this variation in FWS allocation provides promising data to instrument for variation in student employment. Data Integrated Postsecondary Education Data System (IPEDS) Data: The IPEDS contains yearly updated data on all institutions that participate in any federal student financial aid program authorized by Title IV of the Higher Education Act of 1965, amounting to over 7,500 schools (Integrated Postsecondary Education Data System 2013). IPEDS includes data on institutional characteristics, institutional prices, enrollment, student Larissa Muramoto 6/1/2013 8 financial aid, degrees and certificates conferred, student persistence and success, and institutional human and fiscal resources. This dataset is useful to examine the underlying relationship between postsecondary institutions and the FWS program. Since FWS funds are mostly based on the original allocations in 1965, institutions that were richer in the past are more likely to have larger FWS allocations. IPEDS variables such as admissions rate, percentage of students eligible for a Pell Grant, and average tuition are essential to determine how FWS accessibility varies across institutions. Since the IPEDS is extremely comprehensive, the limitations posed by this dataset are minor. The biggest drawback of IPEDS is that it does not contain comprehensive data on the FWS program in particular. This analysis utilizes the 2011 IPEDS release because it has the most financial data out of all the releases. Beginning Postsecondary Students Longitudinal Study (BPS) Data: The BPS contains data on 16,700 students who were surveyed at the end of their first (2003-04), third (2005-06), and sixth (2008-09) years after entering postsecondary education (Beginning Postsecondary Students 2012). This dataset is ideal to examine the causal effects of the FWS program because it includes variables that can be used to examine both post-college outcomes such as unemployment and job benefits and academic outcomes such as degree attainment and time to degree completion. Additionally, the BPS contains rich variables to control for student ability (GPA, SAT/ACT score), demographics (race, gender, EFC), school and work experiences, and degree attainment. The biggest limitation of the BPS is that it is a relatively new dataset. Thus, it does not include more current variables such as income after graduation and outcomes later in these students’ lives. However, the BPS does include some data on reported post-graduation outcomes Larissa Muramoto 6/1/2013 9 such as “consider first job as the start of career” and “job provides good benefits” that can replace income in a regression. While the BPS does not have employment data 3+ years after graduation, these later income statistics might not be interesting because a person’s collegiate experience has the largest impact in the first few years after entering the work force. Pell Grant Data: Every year, the U.S. Department of Education publishes basic data on Pell Grants, including the number of recipients per institution and the monetary amount of awards per institution (Federal Pell Grant Program End-of-Year Report 2006). The Pell Grant dataset includes variables such as institution name, institution state, institution type, number of recipients/institution, and monetary amount of awards/institution. This analysis utilizes Pell Grant data from 2006-2007 because the BPS tracks students from 2004-2009 and 2006-2007 is in the middle of this date range. The Pell Grant data is used construct instruments for FWS participation. Federal Work-Study Data: Similar to the Federal Aid Data, the U.S. Department of Education publishes basic data on the allocation and distribution of work-study funds (Federal Work-Study Program 2011). The FWS dataset includes variables such as institution name, institution state, institution type, number of recipients/institution, federal allocation of Work-Study funds, and number of recipients. The Federal Work-Study data is used to construct instruments for FWS participation. US Census Data: Larissa Muramoto 6/1/2013 10 The US Census contains employment data on work sectors at the country, state, and county level. Variables in the utilized dataset include county, employment, and size of sector. This US Census data is used to construct an instrument for FWS participation that accounts for student employment opportunities around the colleges they attend. Bureau of Labor Statistics (BLS) Employment Data: The BLS provides statistics on the national, state, and county level on occupational employment and wages by age and NAICS super-sector. This data is used to determine the sectors that are most likely to employ college students. This information is essential to determine the sectors to examine in US Census data. Methodology To determine the effect of a FWS job on academic and post-graduate outcomes, one would need to determine the effect of working in college, specifically comparing students with a FWS jobs to those with a non-FWS job. This will isolate the impact of a FWS job over a nonFWS job. I will examine 6 major outcomes of interest: degree completion, time to degree completion, 3 proxies for post-graduate job quality, and unemployment 3+ months after graduation. The proxies for post-graduate job quality are as follows: Received BA/BS, Months to BA/BS, “Offers freedom for other interests,” “Pays the bills”, “Provides benefits”, and Unemployed 3mo After Graduation. Since the BPS does not currently contain wage or earnings data for the students in 2009, these proxies can be used to determine whether FWS helps students secure “quality” jobs. I included unemployment data as well to determine whether participation in the FWS Program improves employment prospects after graduation. Some of the control variables Larissa Muramoto 6/1/2013 11 I included are: high school GPA, SAT/ACT score, gender, race/ethnicity, and EFC. A complete list of control variables and their reason for inclusion is in Appendix A. 2SLS Model First stage equations: ℎ𝑜𝑢𝑟𝑠𝐹𝑊𝑆𝑤𝑜𝑟𝑘! = 𝛼! + 𝛼! 𝑐𝑜𝑙𝑙𝑒𝑔𝑒𝐹𝑊𝑆𝑎𝑣𝑎𝑖𝑙𝑎𝑏𝑖𝑙𝑖𝑡𝑦! + 𝛼! 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐹𝑊𝑆𝑎𝑣𝑎𝑖𝑙𝑎𝑏𝑖𝑙𝑖𝑡𝑦! + 𝒄𝒐𝒏𝒕𝒓𝒐𝒍𝒔𝒊 𝛼! + 𝜐! ℎ𝑜𝑢𝑟𝑠𝑤𝑜𝑟𝑘! = 𝛿! + 𝛿! 𝑐𝑜𝑙𝑙𝑒𝑔𝑒𝐹𝑊𝑆𝑎𝑣𝑎𝑖𝑙𝑎𝑏𝑖𝑙𝑖𝑡𝑦! + 𝛿! 𝑟𝑒𝑙𝑎𝑡𝑖𝑣𝑒𝐹𝑊𝑆𝑎𝑣𝑎𝑖𝑙𝑎𝑏𝑖𝑙𝑖𝑡𝑦! + 𝒄𝒐𝒏𝒕𝒓𝒐𝒍𝒔𝒊 𝛿! + 𝜉! Second stage equation: 𝑦! = 𝛽! + 𝛽! ℎ𝑜𝑢𝑟𝑠𝐹𝑊𝑆𝑤𝑜𝑟𝑘! + 𝛽! ℎ𝑜𝑢𝑟𝑠𝑤𝑜𝑟𝑘! + 𝒄𝒐𝒏𝒕𝒓𝒐𝒍𝒔𝒊 𝛽! + 𝜀! yi represents one of the 6 outcomes of interest: Received BA/BS, Months to BA/BS, “Offers freedom for other interests,” “Pays the bills”, “Provides benefits”, and Unemployed 3mo After Graduation. hoursFWSworki is the number of hours a student worked in a FWS job weekly. hoursworki is the number of hours a student worked weekly. This regression requires 2 instruments because there are 2 endogenous variables, “weekly FWS hours worked” and “weekly hours worked”. Weekly FWS hours worked might be endogenous because selection into the FWS Program is not random. Only individuals who meet Larissa Muramoto 6/1/2013 12 the financial need requirements are eligible to participate in the FWS, and even then, not all FWS-eligible students decide to participate. Additionally, FWS jobs might have restrictions on the number of hours a student can work. Similarly, weekly hours worked might be endogenous because people do not choose the number of hours they work completely randomly. For example, weekly hours worked could correspond with the field a student’s job is in, data I did not have for this analysis. To address these issues, I created 2 instruments: “College FWS Availability” and “Relative FWS Job Availability”. College FWS Availability is defined as: (Allocated FWS Funding)/(Number of Pell Grant Eligible students). I used the FWS and Pell Grant data to construct this instrument. Similar to the instrument that Scott-Clayton employed to examine the effect of the FWS Program in West Virginia, this instrument exploits the variation in FWS allocations that arises because of the way the US government allots FWS funds to students. “College FWS Availability” fulfills condition 1 because college FWS availability is strongly correlated with a person’s participation in the FWS Program. If a college has more federal FWS funding relative to the number of students who are eligible for the funding, a student is more likely to participate in FWS. College FWS Availability also fulfills condition 2 because college FWS availability should not be correlated with a student’s EFC or the number of hours a student works in a college job. While one might argue that students of different income levels might choose schools with different FWS availability, it is unlikely because students often do not choose schools based on their propensity to get a FWS job. In fact, many students are unaware of the FWS availability on campus simply because they do not have access to a school’s job postings prior to enrollment. Additionally, it is unlikely that FWS availability is correlated Larissa Muramoto 6/1/2013 13 with the number of hours a student works in a college job because FWS jobs are relatively standardized across schools. The second instrument, Relative FWS Job Availability is defined as: (Number of FWS jobs)/(Number of jobs around the school). I used the FWS, Census, and BLS employment data to construct this instrument. The FWS data provided the exact number of FWS jobs that were available at each institution. The Census data provided the number of jobs in the county around each institution by NAICS sector. To exclude jobs that would be inaccessible to college students, I used BLS data to determine the supersectors that attract the most college-age workers: wholesale and retail trade, educational and health services, and leisure and hospitality (“Labor Force Statistics” 2013). These supersectors contain the following sectors: wholesale trade (NAICS 42), retail trade (NAICS 44-45), educational services (NAICS 61), health care and social assistance (NAICS 62), arts, entertainment, and recreation (NAICS 71), and accommodation and food services (NAICS 72). Together, these sectors employ 80% of all college-age workers. While employment data was available for zip codes, I decided to examine employment at the county level to account for students’ ability to commute to work. “Relative FWS Job Availability” fulfills condition 1 because relative FWS job availability is strongly correlated a person’s participation in the FWS Program. If there are more FWS jobs relative to the number of comparable, entry-level jobs around the school, a student is more likely to participate in FWS. Additionally, Relative FWS Job Availability fulfills condition 2 because relative FWS job availability should not be correlated with a student’s family income or the number of hours a student works in a college job. It is unlikely that students of different income levels choose schools based on the types of college jobs available around it. Relative Larissa Muramoto 6/1/2013 14 FWS job availability also should not be correlated with the number of hours a student works in a college job. Descriptive Statistics College Students and Employment: In Table 1a, we can see that around 80% of students hold jobs in college, with over half of these students holding a FWS job. Table 1b shows the number of STEM units, a proxy for major, that students take given their employment status. This is important because students with STEM majors often have higher rates of employment and incomes post-graduation. We can see that working in a FWS job corresponds to taking fewer STEM units. Table 2 compares the average hours worked/week and average income/year for students with FWS and alternate jobs. We can see that students with only FWS jobs work fewer hours than those with alternate jobs. However, those students with both a FWS and alternate job work about the same number of hours as those who only work an alternate job. This result is not surprising, given caps on the number of hours a FWS student can work. By comparing students who work multiple jobs, we can see that those who have a FWS job do earn about $1000 less per year than their peers. Postsecondary Institutions, College FWS Availability, and Relative FWS Job Availability: Tables 3 and 4 shows the distribution in my IVs, “College Federal Work-Study Availability” and “Relative Federal Work-Study Job Availability”. The distribution statistics of Larissa Muramoto 6/1/2013 15 College FWS Availability suggest that the distribution is skewed to the right, indicating that there are many schools that have little FWS availability: the median is 13.31, while the mean is 273.53. However, we can also see that College FWS Availability still has variation amongst the remaining schools: the standard deviation is 5,063.57 and the range is 673,974. Relative FWS Job Availability shows a similar right-skewed distribution. While the mean is 0.00225, the median is 0.00003. At the 25th percentile, Relative FWS Job Availability is 0, similar to the College FWS Availability distribution. At the same time, Relative FWS Job Availability has a standard deviation of 0.04630, which is large compared to its mean of 0.00225. Although there are many schools with little FWS availability, there is still variation in FWS availability amongst the remaining schools. To analyze the distribution of College FWS Availability and Relative FWS Job Availability in more detail, I analyzed how these two IVs vary across schools with different admissions rates, percentage of Pell Grant eligible students, and average net price. Tables 4a-c show the average College FWS Availability and Relative FWS Job Availability for a given decile of admissions rates, percentage of Pell Grant eligible students, and net price. We can see from these tables that College FWS Availability is significantly higher in the schools with the lowest admissions rates and lowest percentage of Pell Grant eligible students, perhaps reflecting the result of a legacy-based allocation system where richer, and subsequently more selective schools secure more funds than their poorer counterparts. Additionally, College FWS Availability is higher in the schools that have higher average net prices, reflecting a similar trend. While these trends might not be optimal because schools should be equivalent in all aspects except FWS Availability, one must note that there are still many deciles that look similar Larissa Muramoto 6/1/2013 16 in terms of College FWS Availability. A more detailed analysis of the distribution of College FWS Availability across colleges with different characteristics can be seen in Tables 5a-c. On the other hand, Relative FWS Job Availability does not vary much across schools with different admissions rates, percentage of Pell Grant eligible students, and average net price. This is not surprising because there are many factors besides institution age or affluence that determine the number of jobs available to college students around a particular college. A more detailed analysis of the distribution of Relative FWS Job Availability across colleges with different characteristics can be seen in Tables 6a-c. Results Table 7a shows the result of the six regressions used to examine the impact of a FWS job compared to a non-FWS job, not controlling for a student’s major. Table 7b shows the result of the same regressions used to examine the impact of a FWS job compared to a non-FWS job, controlling for a student’s major. Standard errors are provided in parentheses. Coefficients that are statistically significant at the 5% level are indicated with one asterisk. Coefficients that are statistically significant at the 1% level are indicated with two asterisks. Coefficients that are statistically significant at the 0.1% level are indicated with three asterisks. From the results in Table 7a, we can see that working in an FWS job is correlated with a 5.54% increase in the probability of receiving a bachelor’s degree, a statistically significant result at the 0.1% level. On the other hand, simply working is correlated with a 0.18% decrease in the probability of receiving a bachelor’s degree. While this is a small decrease, the result is Larissa Muramoto 6/1/2013 17 significant at the 1% level. These findings suggest that FWS has a positive impact on academic outcomes while working in general has a slightly negative impact. An explanation for this outcome is that the benefits of FWS jobs, such as helping students feel more integrated on campus, outweigh the costs of having a job such as having less time for studying. This might explain the observed slight decrease in the probability of receiving a bachelor’s degree for nonFWS jobs but a positive effect for FWS jobs. Similarly, working in a FWS job has no effect on the time to graduate, but each hour working non-FWS job is correlated with a 1.87-month increase in the time to graduate, a significant result at the 0.1% level. Since FWS jobs are designed to compliment a student’s education, it is possible that they have a smaller negative effect on a student’s academics than comparable non-FWS jobs. Perhaps FWS jobs give students more flexible working hours and/or require less commute time than non-FWS jobs, making it easier for them to work and study concurrently. Given that majority of students choose to work in college according to the BPS, these results suggest that a student who wants to work in college should take a FWS job if he/she has the opportunity. Additionally, we can see that working in a FWS job has a significant effect on postgraduation outcomes. For example, working an additional hour in a FWS job corresponds to a 12.8% increase in the probability of reporting that one’s first post-graduate job “offers freedom for other interests,” an 8.63% increase in the probability of reporting that it “pays the bills”, and a 10.6% increase in the probability of reporting that it “provides benefits”. These are all indicators of a “better quality” or “more competitive” job. On the other hand, working an additional hour in general only corresponds to a 0.98% increase in the probability of reporting Larissa Muramoto 6/1/2013 18 that one’s first post-graduate job “offers freedom for other interests,” a 3.04% increase in the probability of reporting that it “pays the bills”, and a 0.75% increase in the probability of reporting that it “provides benefits”. These results suggest that working in college is intrinsically beneficial, given that one can find a job post-graduation. One can expect that working during college could help a student obtain workforce experience that he/she could leverage in the job search. However, working in a FWS job seems to provide additional benefits over simply working, suggesting that the skills learned in FWS jobs might translate better to future employment. When controlling for a student’s major, the benefit of working in a FWS job is smaller, yet still larger than the effect of working in a non-FWS job, as can be seen in Table 7b. The probability of reporting that one’s first post-graduate job “offers freedom for other interests” drops to 10.8%, the probability of reporting that it “pays the bills” drops to 5.08%, and the probability of reporting that it “provides benefits” drops to 8.65%. While the estimate on “pays the bills” is loses its significance, the other two estimates are significant at the 5% level. The estimates for the benefit of working in a non-FWS job do not change. These differences might stem from the fact that students of different majors are not equally likely to be employed in a FWS job. Using the number of STEM credits completed as a proxy for major, we can see that working in a FWS job negatively correlates with STEM credits completed. Perhaps because post-graduation unemployment rates are higher for non-STEM majors, those students who do find jobs are more likely to report being satisfied with their employment outcomes than STEM majors. If true, these facts would bias the effect of FWS on job quality outcomes upward if one does not control for major. Larissa Muramoto 6/1/2013 19 While we see positive effects of FWS and working on post-graduation job quality, we paradoxically see negative effects for working on post-graduate unemployment. Working in a FWS job corresponds to a 1.57% increase in the probability of unemployment 3 months after graduation while working in general corresponds to a 0.81% increase in the probability of unemployment. Controlling for a student’s major, we can see that students who work in FWS jobs are 4.51% less likely to report being unemployed 3 months after graduation while students who work in general are 0.91% more likely to report being unemployed. These results seem counterintuitive because prior employment experience should help students find post-graduate jobs more easily. Additionally, graduates who worked in college might also be able to continue as full-time employees of their companies, an advantage they possess over their non-working peers. However, it is also possible that working in college can make students appear to be weaker job candidates upon graduation. From the results discussed earlier, we can see that working in general corresponds to a decrease in the probability of receiving a bachelor’s degree and an increase in the time to degree completion. The effect reversal of working in an FWS job when controlling for major could be explained by the fact that STEM majors are less likely to be employed in a FWS job. Since working in a non-STEM job is correlated with higher rates of unemployment and FWS jobs employ higher percentages of non-STEM majors, not controlling for major in a regression will underestimate the benefits of a FWS job. Overall, these results suggest that working in a FWS job has additional benefits to simply working in college. It was hypothesized that FWS jobs could help students in their job searches by providing them with more opportunities to build workforce-relevant skills while in college than non-FWS jobs. This is plausible because many FWS jobs focus transferrable office skills Larissa Muramoto 6/1/2013 20 such as computer skills and communicating, while non-FWS jobs build less-transferrable skills such as waitressing and manual labor. Perhaps students employed in FWS jobs fare better immediately post-graduation because FWS jobs emphasize skills that are more applicable to jobs that one would consider to be a part of a “permanent career”. Conclusion In conclusion, FWS jobs do indeed have different effects from non-FWS jobs. There is an additional benefit from working more hours in a FWS job: the likelihood of graduating with a bachelor’s degree increases, while it decreases from working in general. Additionally, having a FWS job has no impact on time to graduation, while a non-FWS job does. These results might stem from the FWS Program’s original intention to provide jobs that compliment a student’s education. Alternate jobs commonly taken by college students, such as waitering, might not compliment a student’s education as well. This data also suggests that holding a FWS job in college impacts a student’s postgraduate outcomes. Working in a FWS job is correlated with a decline in the probability of postgraduation unemployment. Additionally, working in a FWS job is correlated with higher likelihoods of reporting that one’s first post-graduate job “offers freedom for other interests,” “pays the bills”, and “provides benefits”: all indicators of “better quality” jobs. While working in general is also correlated with the same indicators, the effect of a FWS job is stronger. Even when controlling for a student’s major, students who were employed in FWS jobs seem to fare better. Larissa Muramoto 6/1/2013 21 However, it must be noted that the students in the analyzed cohort graduated in 2009, an extremely poor job market. The aforementioned effects might disappear in a stronger job market. Given that the unemployment rate in June 2009 was 9.5%, this analysis should be repeated for cohorts graduating in different economic conditions. Larissa Muramoto 6/1/2013 Appendix A: Tables and Figures Table 1a: Students and Work Table 1b: FWS Participation and Number of STEM Credits (Proxy for Major) 22 Larissa Muramoto 6/1/2013 Table 2: Average Hours Worked/Week and Average Monthly Earnings 23 Larissa Muramoto 6/1/2013 Table 3: College Federal Work-Study Availability and Relative FWS Job Availability 24 Larissa Muramoto 6/1/2013 Tables 4a-c: College Federal Work-Study Availability by Admissions Rate, % of Pell-Grant Eligible Matriculated Students, and Net Cost 25 Larissa Muramoto 6/1/2013 26 Larissa Muramoto 6/1/2013 Table 5a-c: Pairwise Comparisons of Means for College Federal Work-Study Availability, across Admissions Rate, % of Pell-Grant Eligible Matriculated Students, and Net Cost 27 Larissa Muramoto 6/1/2013 28 Larissa Muramoto 6/1/2013 29 Tables 6a-c: Pairwise Comparisons of Means for Relative Federal Work-Study Job Availability, across Admissions Rate, % of Pell-Grant Eligible Matriculated Students, and Net Cost Larissa Muramoto 6/1/2013 30 Larissa Muramoto 6/1/2013 31 Table 7a: Hours Worked Weekly and Outcomes, not Controlling for Major (Received BA/BS, Months to BA/BS, “Offers freedom for other interests,” “Pays the bills”, “Provides benefits”, Unemployed 3mo After Graduation) BA/BS Mo. to BA/BS Job Quality Indicators Unemployed "Time for other interests" "Pays bills" "Benefits" 0.174 0.128* 0.0863* 0.106* 0.0157*** (0.42) (2.37) (1.99) (2.3) (5.54) Hours FWS Work 0.0554*** (20.87) Hours Work -0.00180** (-3.09) 1.872*** (33.7) 0.00975** (3.24) 0.0304*** (12.57) 0.00745** (2.91) 0.00814*** (10.93) 10657 5045 1587 1587 1587 5763 N T-statistics are in parentheses * p<0.05, ** p<0.01, *** p<0.001 Table 7b: Hours Worked Weekly and Outcomes, Controlling for Major (Received BA/BS, Months to BA/BS, “Offers freedom for other interests,” “Pays the bills”, “Provides benefits”, Unemployed 3mo After Graduation) BA/BS Hours FWS Work Hours Work N Mo. to BA/BS 0.0582*** (21.83) -0.00347*** (-5.81) 1.828*** (32.59) 0.00995*** (3.67) 10657 5045 1587 T-statistics are in parentheses * p<0.05, ** p<0.01, *** p<0.001 Job Quality Indicators Unemployed "Time for other interests" "Pays bills" "Benefits" 0.140 0.108* 0.0508 0.0865* -0.0451* (0.35) (2.08) (1.35) (1.97) (-1.98) 0.0304*** 0.00757*** (15.46) (3.31) 1587 1587 0.00913*** (11.75) 5763 Larissa Muramoto 6/1/2013 Appendix B: Controls and reasons for their inclusion Control for ability/motivation: • High school GPA • High school GPA squared • ACT test score (or SAT equivalent) • ACT test score (or SAT equivalent) squared • Number of STEM credits completed Control for demographics: • Indicator for female students • Indicator for different races Control for socioeconomic status: • EFC Control for type of school attended: • Type of institution 32 Larissa Muramoto 6/1/2013 33 Appendix C: Reference List Beginning Postsecondary Students (BPS) - Overview. 2012. http://nces.ed.gov/surveys/bps/, accessed May 15, 2013. Davis-Kean, P. E. (2005). The influence of parent education and family income on child achievement: the indirect role of parental expectations and the home environment. Journal of Family Psychology, 19(2), 294. DeSimone J. (2008). The impact of employment during school on college student academic performance (NBER Working Paper No. 14006). Cambridge, MA: National Bureau of Economic Research. Federal Campus-Based Programs Data Book 2012. 2012. Last updated 7/2012. http://www2.ed.gov/finaid/prof/resources/data/databook2012/databook2012.html, accessed May 15, 2013. Federal Student Aid Handbook. 2003. Last updated 5/2003. http://www.ifap.ed.gov/fsahandbook/0304FSAHBVol6FWS.html, accessed May 15, 2013. Federal Pell Grant Program End-of-Year Report. 2006. Last updated 5/2008. http://www2.ed.gov/finaid/prof/resources/data/pell-2006-07/pell-eoy-2006-07.html, accessed May 15, 2013. Federal Work-Study (FWS) Program. 2011. Last updated 12/2011. http://www2.ed.gov/programs/fws/index.html, accessed May 15, 2013. Hossler, D., Ziskin, M., Gross, J. P., Kim, S., & Cekic, O. (2009). Student aid and its role in encouraging persistence. Higher education: Handbook of theory and research, 389-425. Larissa Muramoto 6/1/2013 34 Integrated Postsecondary Education Data System. 2013. Last updated 5/2013. http://nces.ed.gov/ipeds/, accessed May 15, 2013. Kalenkoski C. M., Pabilonia S. W.(2010). Parental transfers, student achievement, and the labor supply of college students. Journal of Population Economics, 23, 469–496. Labor Force Statistics from the Current Population Survey. 2013. “Employed persons in nonagricultural industries by age, sex, race, and Hispanic or Latino ethnicity.” Last updated 2/2013. http://www.bls.gov/cps/cpsaat14.htm, accessed May 15, 2013. Scott-Clayton, J. (2011). The Causal Effect of Federal Work-Study Participation QuasiExperimental Evidence From West Virginia. Educational Evaluation and Policy Analysis, 33(4), 506-527. Smole D. P. (2005). The campus-based financial aid programs: A review and analysis of the allocation of funds to institutions and the distribution of aid to students (Congressional Research Service Report RL 32775). Washington, DC:Congressional Research Service. Stinebrickner R., Stinebrickner T. R. (2003). Working during school and academic performance. Journal of Labor Economics 21, 473–491. Tinto, V. (1987). Leaving college: Rethinking the causes and cures of student attrition. University of Chicago Press, 5801 S. Ellis Avenue, Chicago, IL 60637.