The Impact of Education-Job Match paper

The Impact of Education-Job Match on Earnings and Job Satisfaction: A Comparison of STEM and Non-STEM Workers Ashlee Frandell, Eric Welch, Chris Herbst Abstract: Substantial federal and state funding has been invested in fostering the Science, Technology, Engineering and Mathematics (STEM) workforce in the US with the expectation that university graduates at all levels will be able to obtain jobs that make the most of their learning and skills Yet the extent to which STEM graduates are able to obtain jobs in their fields and put their learning to use is not clear Some reports claim that the supply of STEM workers does not meet market demand, while others find there is a glut of STEM workers One way to look at this complexity is to examine education-job match, which is one measure of graduates’ ability to use the knowledge and skills acquired at college in the labor market Match occurs when workers’ jobs correspond well with their educational fields, which has implications for the supply and demand of STEM workers as well as returns to their education This study asks: 1) Is there more education-job match for STEM workers compared to non-STEM workers? 2) How does the match affect income and job satisfaction? Results show that there are differences in educationjob match for STEM and non-STEM workers for the full sample and across all degree levels Also, the level of education-job match positively affects workers’ returns to education in terms of income and overall job satisfaction, for some degree levels but not others This study will contribute to the knowledge base on the STEM workforce and the impact of education-job match level on labor market outcomes We use a more comprehensive approach that has methodological advantages and will add to the body of literature Keywords: Education-Job Match, STEM workforce, income, job satisfaction 1 Introduction Significant federal and state funding has been invested in building the Science, Technology, Engineering and Mathematics (STEM) workforce in the US with the expectation that university graduates at all levels will be able to obtain jobs that make the most of their learning and skills Most agree that technological changes are creating new jobs and fields and that on average STEM graduates earn more than non-STEM graduates (White, et al, 2019) Yet the extent to which STEM graduates are able to obtain jobs in their fields and put their learning to use is not clear Some reports claim that the supply of STEM workers does not meet market demand and that employers struggle to fill STEM specific positions, while others find there is a glut of STEM workers (White, et al, 2019) In part, the complexity of the STEM workforce – as defined by discipline, degree, cohort, etc - makes it difficult to quantify STEM job availability and compensation One way to look at this complexity is to examine the degree of education-job match among workers in STEM fields compared to those in non-STEM employment Education-job match is one measure of graduates’ ability to use the knowledge and skills acquired at college in the labor market (Hur, et al., 2019) Match occurs when workers’ jobs correspond with their educational fields and the level depends on the type of schooling received and how well it corresponds to the workers’ occupations (Montt 2017; Robst 2008; Quintini 2011a) The level of education-job match can influence the supply and demand of STEM workers as well as their returns to education A low level of match or mismatch is one indicator of surplus or deficit human capital in terms of the overeducation or under-skilling of STEM workers, which can adversely influence labor market outcomes (McGuinness, et al, 2017) For instance, mismatched workers tend to earn less and have lower job satisfaction compared to their counterparts in positions more aligned with their degree field, which harms worker productivity and retention (Hur, et al., 2019) Lower job satisfaction also has been shown to decrease productivity and is one reason why mismatched workers search for new jobs (Allen & Van der Velden, 2001; Kim and Oh, 2002) Although earlier research has found that education-job match also varies by field of study, far less work has looked at this type of match for STEM workers in particular Additionally, none of these studies directly compare the education-job match for STEM and non-STEM employees Therefore, this study asks: 1) Is there more education-job match for STEM workers compared to non-STEM workers across degree levels? and 2) How does the match level affect income and job satisfaction? The rest of the paper is structured as follows Section reviews the literature on education-job match Section describes the data and summary statistics Section explains the methodology and estimates of the factors affecting match, income and job satisfaction Sections presents the conclusion and discussion Prior Research The literature on education-job match and mismatch generally looks at predictors of match and the impact of mismatch on labor market outcomes Though, one gap in the literature is that the studies not specify the importance of some of the key predictors or controls Examples of such factors that may matter over time are field of study and level of education These covariates can have important implications for match and its relationship to the labor market Thus, the aim of this paper is to better understand the importance of STEM versus non-STEM classification and level of education for level of match and its effect on other outcomes Some studies analyze education-job match for multiple majors with no particular attention to STEM fields Robst (2007) examines education-job match level and finds that almost all majors have a higher chance of mismatch compared to computer and information sciences The highest rates are for English, foreign languages, social sciences, and liberal arts majors This shows that majors that typically provide more general skills, as opposed to occupation-specific skills tend to have lower rates of match (Robst, 2007) An analysis by Sellami and colleagues (2018) confirms this by showing that graduates with a degree in Arts followed by Linguistics, History or Philosophy typically have a lower probability of match According to the authors, fewer job openings for these graduates means they must take jobs outside of their field of study Other predictors included in their mismatch models are academic performance and educational level (Sellami, Verhaest and Trier, 2018) A study of the determinants of match for Canadian universities also shows differences in the level of education-job match between educational programs and level of education (Boudarbat and Chernoff, 2010) While these studies control for the workers’ field of study, there are many comparison groups due to the large range of fields included, which restricts the implications and generalizability Studies also find that the level of education-job match influences graduates’ success in the labor market and impacts their return on the investment in their education The findings of Bender and Heywood (2009) and Robst (2008, 2007a) imply that low job match is associated with reduced earnings and job satisfaction Using field of study fixed effects as controls, Rios-Avila and Saavedra-Caballero (2019) find a three to four percent wage boost for employees that have good job matches Workers that take jobs that are not well matched to the skills they gained in college tend to yield lower wages, as their capabilities are not being fully employed (Béduwé & Giret, 2011) Other studies confirm that match matters for labor market outcomes including wages and satisfaction, and that a combination of education field and overeducation influence the match (Beduwe and Giret, 2011; Støren and Arnesen, 2011; Wolbers, 2003; Xu, 2013) Liwiński and Pastore (2019) find both a positive correlation of job matching with earnings and a positive wage premium from education This study expects to find that a high level of education-job match positively affects job earnings and satisfaction A relatively small literature focuses on STEM rather than all majors in order to examine match and its impact on outcomes However, most of these studies not include non-STEM comparison fields In contrast, this study looks at how the outcomes vary by degree classification of STEM versus non-STEM The STEM focused study by Xue & Larson (2015) looks at matches and mismatches between the supply and demand of STEM workers Their results show that the academic sector is oversupplied while the private and government sectors have specific shortages for certain fields such as for data scientists, and software developers Alhaddab (2015) tests predictors and measures the level of STEM degree-job match The study finds that mismatched workers have about a 33% decrease in wages compared to their counterparts This is consistent with the labor market outcomes of the broader multi-field match studies Bender & Heywood (2009) also find that STEM workers suffer due to low match, but that they are disadvantaged differently based on their skillsets The authors maintain that STEM fields generally require higher skills that change more rapidly than in other fields, which leads to greater mismatch and associated penalties for scientists However, this conclusion depends on many factors including discipline, and technological and workforce changes (Bender & Heywood, 2009) Based on the STEM and other literature, we expect the academic field will significantly impact labor market outcomes Specifically, we expect the level of degree-job match and its influence on earnings and job satisfaction to differ for STEM and non-STEM workers Another important covariate from the literature that influences labor market outcomes is education level Larson, et al (2014) find that the matching of STEM graduates to jobs varies by both field specialization and education level The study argues that there are key differences between the labor market for PhD students and the one for students with lower level STEM credentials The findings indicate that there are more PhD students than the academic market needs in many of the subfields While the academic market is oversupplied, there is still a need for STEM workers with undergraduate or master’s degrees (Larson, et al 2014) Xue & Larson (2015) find that the demand and supply of STEM workers differ by degree specialization across academic levels They also find that there is a surplus of workers with PhDs, mostly in academic tenure-track positions and in certain fields in industry and government In comparison, there is more consistent demand for employees with bachelor’s and master’s degrees, also differing by field specialization (Xue & Larson, 2015) Again, these studies mainly look at differences in outcomes for STEM workers by degree level without comparing them to the outcomes of workers in non-STEM fields Other studies focus on the academic and other markets for PhD graduates and postdocs, and find distinct outcomes for this higher education level Spronken-Smith (2018) discusses the growth in the number of PhD graduates compared to the availability of faculty jobs Due to this and the opportunities beyond academia, roughly half of the PhD graduates find positions outside of universities (Guthrie & Bryant, 2015; McAlpine, 2016; Mellors-Bourne et al., 2013; Neumann & Tan, 2011; Spronken-Smith, 2018) Some of the literature finds differences in the career aspirations between PhD students in STEM fields and PhD students in other academic fields Students in the certain STEM fields with strong ties to industry are less likely to seek academic jobs from the start of their PhD programs compared to students in other fields (Barry 2013; Golde and Dore 2004, Etmanski, 2019) Discrepancies between expected and actual career outcomes based on educational field and level can lower the satisfaction rate for these PhD graduates (Spronken-Smith, 2018) These findings imply that differences between STEM and non-STEM workers may be unique to this subpopulation of higher-level graduates, and so outcomes may differ for undergraduates One study by Rose (2017) focuses on whether undergraduates with bachelor’s degrees are employed in good-fitting jobs The results include that in 2014, 25% of these workers were overqualified for their jobs and that their income penalty increased over time starting in the 1980s to reach 50% Though these studies not treat workers with different levels of education as separate subpopulations, the findings show consistent differences for the levels in terms of degree-job match, earnings and satisfaction (Hur, et al, 2019; Bol, 2019; Boudarbat and Chernoff, 2010) For instance, Hur, et al., (2019) finds that education-job match increases for employees with a higher degree Workers with bachelor’s degrees have the lowest match followed by master’s and then doctoral graduates Thus, this study will treat individuals with different levels of education as separate subpopulations to more accurately analyze differences in match as well as the earnings and job satisfaction for STEM and non-STEM workers We expect to find significant differences for workers with graduate and undergraduate degrees While Hur, et al., (2019) is an important study that finds differences in match, it along with much of the education-job match literature does not directly compare STEM and non-STEM workers or separate subpopulations by degree level (bachelors, masters and doctoral) The present study intends to fill this gap and complement Hur, et al., (2019) by analyzing the level of match and its relationship to earnings and job satisfaction for STEM and non-STEM workers by degree level over time By doing so, we aim to better analyze job mismatch by these key covariates (degree type and level) and to expand the knowledge on the state of the STEM labor market and outcomes for each degree level Additionally, Hur, et al.’s (2019) estimation strategy does not account for the unobserved determinants of earnings and job satisfaction that may be correlated with match As a result, estimates of the impact of match on earnings and job satisfaction may be biased This study, in contrast, attempts to recover the causal effect of match by controlling for individual fixed effects, which account for all unobserved time invariant worker characteristics that my influence earnings and satisfaction Data The longitudinal data for this study was obtained through the Integrated Public Use Microdata Series (IPUMS) HIGHER ED IPUMS integrates three National Science Foundation (NSF) surveys: the National Surveys of College Graduates (NSCG), the National Survey of Recent College Graduates (NSRCG) and the Survey of Doctorate Recipients (SDR) The instruments of all three questionnaires are the same over time as the surveys are designed to allow for follow-up sampling and to reach a large target population The database matches individuals across surveys over time using two variables (PERSONID and REFID) and includes statistical weights so that individuals who were eligible for more than one survey are not overrepresented1 It also ensures that the respondents are more representative of the greater U.S population (Minnesota Population Center IPUMS) Thus our analytic sample is a panel of individuals over the years 2003 to 2013, which provides a methodological advantage that allows for stronger modeling The dataset includes information on demographics, education, employment, and income The population target is those under the age of 76 with at least a bachelor’s degree Because the intent of this paper is to analyze educationjob match for the STEM and non-STEM labor force in the United States, individuals who are no longer in the labor market are excluded from the analysis for that given year Individuals who begin working the next year are added back into the analysis sample The dataset contains multiple STEM and non-STEM fields This study adheres to the NSF classifications for what fields are considered STEM or non-STEM given that the data originated In the IPUMs datasets, individuals are identified across survey years by the generated PERSONID variable, which replaces the original reference REFID variable from that organization The NSF STEM disciplines include agricultural sciences, chemistry, environmental science, mathematics, engineering and life/biological sciences (NSF - STEM Classification Crosswalk) Thus, the non-STEM disciplines for the purpose of this study are social sciences, economics, psychology, management and administration, and other non-science and engineering Table reports the percentage of workers with STEM and non-STEM degrees across education levels The ranges show that the differences in percentages for STEM and nonSTEM workers is pretty consistent from 2003 to 2013 This study uses the self-reported data contained in IPUMS to measure the level of education-job match For our variables of interest, perception is important As such, we expect the respondents to better know their own degree of match and level of satisfaction as compared to using another type of measure if even available Other studies similarly use the subjective measures for match and satisfaction arguing that they provide more important or relevant information about the survey respondents given the nature of the instruments (Robst, 2007; Sellami, Verhaest and Trier, 2018) Additionally, both subjective and objective measures tend to give consistent findings (Lemieux, 2014) Respondents were asked to rate how related their principal job is to their highest degree The categories are not related, somewhat related, and closely related In order to determine clear differences between education-job match and mismatch, the measure was transformed into a binary dummy variable Individuals that chose “closely related” were coded as closely matched (=1), and those that chose “somewhat related” or “not related” were as not closely matched or mismatched (=0), which is the reference group Coding in this way improves the data as it is difficult to interpret what somewhat related means The percentage over time of all workers who are closely matched is 57% while the percentage for workers that are not closely matched is 43% (Table 1) The percentage of closely matched workers is larger for those with graduate degrees compared to undergraduate degrees Workers with a master’s or professional degree have the highest amount of closely matched workers Graph shows the proportion of closely matched STEM workers by degree level over time Match proportions are lower for undergraduate than graduate degrees Graph shows the difference in closely matched proportions for STEM and non-STEM employees STEM workers have a higher proportion of education-job matches than non-STEM and the full sample of workers Table Percentage of Match and Field Across Degree Levels Percentages (%) Closely matched 56.89 Undergraduate 46.52 Graduate 71.03 MA & Professional 71.42 Doctorate 67.73 Not closely matched 43.11 Undergraduate 53.48 Graduate 28.97 MA & Professional 28.58 Doctorate 32.27 STEM 63.01 Undergraduate 70.06 Graduate 53.41 MA & Professional 51.51 Doctorate 69.40 Non-STEM 36.99 Undergraduate 29.94 Graduate 46.59 MA & Professional 48.49 Doctorate 30.60 Graph Proportion of Closely Matched STEM Workers by Degree Level Over time Graph Proportion of Closely Matched STEM/Non-STEM Workers Over time The respondents reported their annual salaries for each survey year excluding any deductions, overtime or bonuses Our sample is restricted to full-time workers because the data only include a measure for annual income Because IPUMS defines full time work as 40 and 52 hours per week, our sample excludes only 8.41% of all IPUMS data Table reports the mean annual earnings over time of the workers by match and degree type Earnings were higher for STEM workers for each category of job-education mismatch and degree level Regardless of degree type, workers who were closely matched earned more than employees who were not, with the possible exception of non-STEM undergraduates Table Income of Workers by Type of Match, Field and Degree Level Mean Annual Earnings ($) Closely matched STEM 79,688.40 Undergraduate 70,316.26 Graduate 91,745.45 Non-STEM 73,545.02 Undergraduate 55,759.90 Graduate 80,251.30 Not closely matched STEM 66,187.50 Undergraduate 62,844.94 Graduate 77,296.92 Non-STEM 60,857.32 Undergraduate 55,075.18 Graduate 71,527.44 To measure satisfaction, the respondents are asked to rate their overall satisfaction with their job at the time each survey was given The four categories were very satisfied, somewhat satisfied, somewhat dissatisfied, and very dissatisfied This measure was transformed into a binary response variable for simplification purposes so that individuals were either coded as satisfied (=1) if they stated they were very satisfied or somewhat satisfied with their jobs and as dissatisfied (=0) if they stated they were somewhat dissatisfied or very dissatisfied with their jobs Table shows the average rates of overall job satisfaction over time varying by match and degree type Workers in closely matched jobs were more satisfied than their counterparts and, in both cases, STEM employees were slightly more satisfied than those with non-STEM degrees Table Job Satisfaction of Workers by Type of Match, Field and Degree Level Satisfied % Not satisfied % Closely matched STEM Undergraduate Graduate Non-STEM Undergraduate Graduate 93.72 93.21 94.38 92.51 91.72 92.81 6.28 6.79 5.62 7.49 8.28 7.19 STEM Undergraduate Graduate Non-STEM Undergraduate Graduate 86.43 86.23 87.07 85.85 85.45 86.58 13.57 13.77 12.93 14.15 14.55 13.42 Not closely matched Estimation Methodology and Results This study seeks to understand how key predictors affect education-job match and its relationship to earnings and job satisfaction We examine variations in match, earnings, and satisfaction outcomes for STEM and non-STEM workers by degree level over time Due to potential differences by degree levels, all regressions are run for the full sample and then separately for each academic degree level 4.1 Education-Job Match We first examine whether there is more education-job match for STEM workers compared to non-STEM workers In order to estimate match difference, this study performed logistic regressions with individual fixed effects to estimate the following education-job match function: (1) log �� (�=� ) = YitSTEMit Xit + YEARt ai + uit �� ( �=�′ ) Where j is the identified outcome (closely matched) and j' is the reference outcome (not closely matched) STEMit is the dummy variable that equals one if the worker has a degree in a STEM field and zero otherwise Xit is a matrix of all other relevant characteristics that are used as control variables in the estimation and YEARt controls for year fixed effects is the individual fixed effects, and uit is the error term Individual fixed effects control for unobserved timeinvariant individual heterogeneity, thus removing the impact of demographic and other variables such as gender, race, socioeconomic status, cognitive abilities and university quality characteristics For example, even though several studies find that GPA is a good predictor of match because employers perceive it as an indicator of ability (Grayson 2004; Boudarbat and Chernoff, 2010; Garcia-Espejo and Ibanez, 2006; Storen and Arnesen, 2006), the IPUMS dataset does not have data on academic and university quality characteristics The individual fixed effects control for the lack of this information and strengthens the results and also reduces concerns about self-selection and omitted variable biases Thus, using fixed effects produces a more rigorous model than what is done by many of the other match studies, which contributes to the body of literature (Hur, et al., 2019) Model covariates The covariates of interest that are included in the model are age and the employer characteristics The other potential covariates not included due to lack of data or importance are controlled for by employing individual fixed effects The literature on this subject shows mixed results about how age impacts the level of education-job match Krahn & Bowlby (1999) found that older workers are more likely to match while later researchers contradict this finding and showed that younger workers are more likely to match (Robst, 2007; Wolbers, 2003) This could be because workers may change jobs and attain new skills reducing their job education match Also, degree fields and the skills required change over time leading to match differences depending on when the workers graduated (Kuehn & Hecker, 2018; Bauer 2002; Bender and Heywood 2011) This study used age intervals as a control in order to account for age differences and cohort effects Employer characteristics included in the models are the employer sector (2-year college/other school, 4-year college/medical institution, Government, Business or industry), employer size, and if the individual received any job-related training A lot of the literature does include the employer sector as an important component of match and finds differences between academic and other sectors (Lee and Sabharwal, 2016; Xue & Larson, 2015) The research contains mixed results as to the influence of firm size, though this predictor is not included often Witte and Kalleberg (1995) found that larger firms are negatively associated with match for males They hypothesize that this is because larger firms have more chances for career advancement and so the workers are less likely to be matched Alternatively, Wolbers (2003) found that working for larger firms increases the match, hypothesizing that this is due to larger firms having more positions available Experiences and training can impact education-job match and labor market outcomes Allen & De Wert (2007) describe how skills gained from experiences and training beyond education are important and allow workers to perform more complex tasks These skilled workers usually have higher wages Work-related training was included in the models to account for how skills gained outside of academia can impact the outcome variables Studies that include academic achievements find that workers with higher college educational levels are more likely to match than those with only a bachelor’s degree (Wolbers 2003; Robst 2007a; Krahn and Bowlby 1999) Thus, this study runs the models for the full sample of workers and then subdivides the sample 10 by academic degree levels That way the models show clear differences in terms of match and the two labor market outcomes across academic degree levels Table shows the summary statistics of the covariates included in the study across all periods of time For the employer characteristics, around 9% of the individuals work for a two-year college or another school system, 7% work for four-year college or a medical institution, 12% work for the government and 72% work for business or industry It is worth noting that around 65% of the sample attended work related training at some point Regarding the highest certificate or degree earned by the workers, 58% received a bachelor’s degree, 28% received a Master’s, and around 15% received a Professional or Doctorate degree The table also shows the weighted percentage of the different variables by match type, which indicates differences among respondents by predictors The larger differences are for employer characteristics and highest degree earned For instance, workers employed at a two-year college are more represented (12%) among closely matched individuals than those not closely matched or mismatched (5%) This is the same for workers employed at four-year colleges or in the government However, employees employed in business or industry are more represented (78%) among individuals not closely matched Additionally, workers who received work-related training are more represented (72%) by closely matched respondents Further, individuals with only a bachelor’s degree are more represented (72%) by those who are not closely matched In contrast, the other three degreed types are more represented by individuals that are closely matched Table Summary statistics of Independent Variables Across All Time Periods % of total sample % of closely matched 68.48 32.42 % not closely matched 55.78 44.43 STEM Non-STEM 63.00 37.58 Age 23 to 34 Age 35 to 44 Age 45 to 54 Age 55 to 64 Age 65 to 75 25.57 26.95 26.94 17.00 3.42 25.92 27.53 27.20 16.00 3.24 25.11 26.19 26.60 18.32 3.67 8.86 7.22 12.14 71.78 19.87 24.19 22.07 14.23 19.06 65.01 11.78 9.24 11.82 67.16 18.35 24.59 23.88 14.73 18.46 72.16 4.99 4.56 12.57 77.88 21.87 23.67 19.68 13.57 21.21 55.58 57.68 28.23 4.49 9.61 47.16 32.10 5.34 15.40 71.56 23.12 3.36 1.96 Employer characteristics 2-year college/other school 4-year college/medical institution Government Business or industry Less than 25 employees 25-499 employees 500-4999 employees 5000-24999 employees 25000+ employees Attended work-related training Highest certificate/degree Bachelor's Master's Doctorate Professional 11 Findings The output for the full sample (Table 5) shows that degree classification as STEM or non-STEM is statistically significant and positively related to being closely matched This indicates that the odds of being closely matched for workers with a degree in STEM are 1.30 times greater than for those with a non-STEM degree For the subsamples, all of the coefficients across degree levels are statistically significant Both STEM workers with only an undergraduate degree or a doctorate are more likely to be closely matched than their non-STEM counterparts The odds of being matched are 6.16 times greater for STEM workers with an undergraduate degree than for non-STEM workers with an undergraduate degree While the odds are 1.75 times greater for STEM workers with a doctorate than their counterparts In comparison, the odds being closely matched for STEM workers with a masters’ or professional degree are 0.73 times smaller than for those with a non-STEM masters’ or professional degree Regarding the employer sector characteristics, workers in business or industry have the lowest odds of being closely matched compared to the other three categories across all degree levels Compared to the smallest employer size, workers in all other sizes have higher odds of being matched in the full sample and for those with undergraduate degrees Relative to the reference category of 23-34 years old, the odds of match worsen with age, though they are not consistently significant across degree levels For the first three columns, those who received job related training have higher odds of being closely matched compared to those who have not An interesting finding is that the impact of training is not significant for those with a doctorate 12 Table Job-Education Match Model Results (Odds Ratios) (1) (2) Full sample Undergraduates STEM Age 35 to 44 (ref Age 23 to 34) Age 45 to 54 Age 55 to 64 Age 65 to 75 2-year college/other school (ref Industry) 4-year college/medical institution Government 25-499 employees (ref Less than 25 employees) 500-4999 employees 5000-24999 employees 25000+ employees Attended work-related training Doctorate (ref undergraduate) MA & professional 2006 (ref 2003) 2008 2010 2013 Obs Pseudo R2 1.30*** (0.10) 0.99 (0.04) 0.95 (0.05) 0.95 (0.07) 0.94 (0.09) 1.88*** (0.13) 1.85*** (0.08) 1.23*** (0.06) 1.16*** (0.04) 1.13*** (0.04) 1.11** (0.05) 1.18*** (0.05) 1.11*** (0.02) 1.97*** (0.34) 2.49*** (0.17) 6.16** (4.77) 0.99 (0.06) 0.84* (0.08) 0.81 (0.10) 0.67** (0.12) 2.28*** (0.27) 1.59*** (0.16) 1.29*** (0.10) 1.29*** (0.08) 1.33*** (0.09) 1.32*** (0.09) 1.40*** (0.10) 1.12*** (0.03) 0.76*** (0.01) 0.87*** (0.02) 0.82*** (0.02) 0.77*** (0.03) 87704 0.01 0.79*** (0.02) 0.94* (0.03) 0.91** (0.04) 0.87** (0.06) 30214 0.01 Standard errors are in parenthesis *** p