The impact of minimum competency exams and rigorous courses on college attendance and early

341 J.H. Bishop et al. Economics of Education Review 19 2000 333–349 Clearly, New York invests a great deal in its K-12 education system. If the cause of the high spending were high costs of living, a strong general commitment to edu- cation or legislative profligacy, we would expect spend- ing to be high on both K-12 and higher education. This is not the case. New York ranks second in the ratio of K-12 spending per pupil to higher education spending per college student and third in the ratio of K-12 teacher salaries to college faculty salaries NCES, 1998.

3. The impact of minimum competency exams and rigorous courses on college attendance and early

success in the labor market What is the impact of minimum competency exam graduation requirements and taking more rigorous higher standards courses on college attendance rates and labor market outcomes? Many employers complain about the basic skills and work habits of recent high school gradu- ates. A black personnel director interviewed for a CBS special on the educational reform proudly stated, “We don’t hire high school graduates any more, we need skilled workers” 6 September, CBS, 1990. A super- visor at New York Life Insurance, for example, asserted on television “When kids come out of high school, they think the world owes them a living” 27 March, PBS, 1989. Surely these generalizations do not apply to every graduate, but the students who are disciplined and aca- demically well prepared currently find it difficult to sig- nal this fact to employers. Why is it so difficult for high school graduates to sig- nal their competencies to employers? One cause is the fear that many employers have of an EEO lawsuit if tests of basic reading and math skills are a component of their employee selection process. Some employers ask for high school transcripts, but they are often late in arriving and the varying standards across courses make them hard to interpret. Other employers ask applicants to enter their high school GPA on the job applications, but seldom do they check whether the information is accurate. Most employers do not even ask for information on grades. Consequently, in most communities, competencies developed in the local high school are poorly signaled to employers. The lack of signals of achievement in high school tend to make employers with the best jobs reluc- tant to risk hiring recent high school graduates. They pre- fer, instead, to hire workers with many years of work experience because the applicant’s work record serves as a signal of competence and reliability that helps them identify the most qualified. 3.1. Minimum competency exam graduation requirements Establishing a minimum competency exam, therefore, is one way a high school or a state school system can try to overcome this signaling problem and help its graduates get good jobs. The existence of the minimum competency exam MCE graduation requirement is eas- ily and cheaply signaled to all local employers. With the MCE requirement, the school’s diploma now signals more than just seat time; it signals meeting or exceeding certain minimum standards in reading, writing and math- ematics. Graduates will have studied harder and be more competent. The diplomaMCE signal is also superior to a GED because it indicates that the student was able to satisfy her teachers as well as perform on standardized tests. This should make local employers more willing to hire the school’s recent graduates. This logic generates an interesting testable prediction: graduates of high schools with an MCE graduation requirement will earn more than graduates of schools without MCEs, even when socio-economic status SES, test scores, grades, types of courses taken, working dur- ing senior year, current and past college attendance and other individual and school characteristics are held con- stant. This hypothesis will be tested in the two nationally representative longitudinal data sets—the National Edu- cational Longitudinal Study NELS88 and High School and Beyond HSB seniors—that asked students whether an MCE had to be passed to get a high school diploma. 6 The analysis sample are the students in the two longitudi- nal studies who graduated from high school between Jan- uary and September of their scheduled year of gradu- ation. The HSB seniors were interviewed 2, 4 and 6 years after graduating from high school so we are able to assess both short and intermediate run effects of student coursework and school policies. NELS88 graduates were interviewed in 1994, 2 years after they graduated. The results of these analyses are presented in Table 3. Mean and standard deviations of dependent variables are presented in Table 6 of Appendix A. When wage or earn- ings are the dependent variable, months attending college full-time and months attending college part-time both current and past are included as control variables. A large number of control variables listed in the table note were included in all of the estimating equations. MCEs had substantial effects on the earnings of young women averaging 5.1 per year who graduated from high school in 1980. The effect was statistically signifi- cant in 3 of the 5 years 1982, 1984 and 1985 analyzed. MCEs had no effect on the earnings of the male mem- 6 The HSB and NELS-88 question on MCEs asked the stud- ent the following: “Does your high school have a minimum competency or proficiency test—that is a special test that all students must pass in order to get a high school diploma?” The question about remedial education was: “Have you ever been in any of the following kinds of courses or programs? a Remedial English sometimes called basic or essential; b Remedial Mathematics sometimes called basic or essential.” 342 J.H. Bishop et al. Economics of Education Review 19 2000 333–349 Table 3 The impact of minimum competency exam graduation requirement on wages of high school graduates a Females Males Coef. T stat Coef. T Stat. 1992 high school graduates Hourly wage rate in 1993 0.034 1.95 0.038 2.12 Earnings in 1993 0.060 1.82 0.005 0.17 1980 high school graduates Earnings in 1981 0.019 0.71 20.045 1.34 Earnings in 1982 0.074 2.05 20.013 0.40 Earnings in 1983 0.032 0.98 20.022 0.68 Earnings in 1984 0.061 2.07 0.004 0.15 Earnings in 1985 0.071 2.68 0.026 1.12 Source: Analysis of HSB and NELS88 data. a Coefficients of the earnings models have been divided by mean earnings for that year. Controls include past and present college attendance, reading and math test scores in 12th grade, grade point average, courses taken in high school, extra-curricular activities, work for pay during senior year, TV and homework hours, religion, reading for pleasure, attitudes, indicator for being handicapped, family demographics, marital and parental status at the end of 12th grade, dummies for region and rural, suburban and urban resi- dence. p,0.01 on a two tail test; p,0.05 on a two tail test; p,0.10 on a two tail test. bers of the 1980 graduating class. Employer demand for literate employees has increased since 1980. This may have increased the signaling value of an MCE. For the 1992 graduating class, hourly wage rates were signifi- cantly 3.4–3.6 higher where passing an MCE was a graduation requirement. In addition, the earnings of female graduates from MCE high schools were a sig- nificant 6 higher. 7 Since the regression model includes controls for 12th grade test scores, impacts that operate through increased test scores are not being counted in the above estimates. The total effect of MCEs on earn- ings will likely be even larger. If the signaling benefits of an MCE are even just a fraction of these levels, they are likely to outweigh any potential losses from a small increase in dropout rates. 8 7 States and school districts with such exams may be differ- ent along unmeasured dimensions that have direct effects on wage levels. A positive selection bias is unlikely, however, because most states appear to have adopted MCEs as a response to a perception that the state’s schools were failing to teach basic skills. By 1992, MCEs had been adopted by every south- ern state except Arkansas and Louisiana. With the exception of Arizona, none of the Mountain, Plains or Midwestern states had established an MCE prior to 1992 NCES, 1993, p. 149. Bishop and Mane 1998 estimate models with controls for teacher sal- ary, average daily attendance, percent of faculty with Masters degrees, a school problems index and indicator variables for private school or denominationally controlled school and find an even stronger relationship between MCEs and wage rates and earnings for both men and women. 8 Analyses of NELS data by Dean Lillard 1997 and Lillard and DeCicca 1997a,b and Lillard and DeCicca 1997a,b sug- gest that minimum competency exam requirements for high school graduation may effect high school dropout rates. Coef- ficients were generally insignificant and varied with specifi- Theory predicts that the earnings effects of an MCE should depend positively on the rigor of the graduation standard. Requiring students to pass the much more dif- ficult Regents exams should therefore strengthen the sig- naling effect described above. It may, however, increase dropout rates. 3.2. Taking more rigorous courses A second way schools can introduce higher graduation standards is to make it difficult or impossible for students to take courses like general math and remedial English that offer graduation credit for reviewing material sup- posed to have been learned in the first 8 years of school. New York State’s requirement that all students take and pass Regents examinations in five core subjects essen- tially forces all high schools to do just that. What is the likely effect of restricting access to basic and remedial courses? Studies have found that, holding initial ability constant, students who take remedial andor basic courses get higher grades but learn less Meyer, 1992; Bishop, 1996. In this paper we examine the effects of taking remedial courses on post-high school labor market outcomes and college attendance. High school graduates who have taken basic and remedial courses are compared to those who have not. Controls are included for as many confounding factors as possible: SES; ethnicity; 12th grade test scores; GPA; numbers of vocational and aca- demic courses taken; extracurricular activities; hours doing homework; hours watching TV; attitudes and demographic characteristics. Almost identical models cation. Models that controlled for state fixed effects and exam- ine the effect of introducing a state MCE tend to find no effect. 343 J.H. Bishop et al. Economics of Education Review 19 2000 333–349 were estimated in three different nationally representa- tive longitudinal data sets: NELS88, HSB and the National Longitudinal Survey of 1972 NLS72. Results are reported in Table 4. The variable of inter- est takes on a value of 1 if basic or remedial classes have been taken in both English and mathematics. It takes on a value of 0.5 if basic or remedial classes have been taken in just one subject. Participation in such classes grew dramatically between 1972 and 1980 and then fell a little by 1992. The mean value of this variable was 0.06 for the Class of 1972, 0.29 for the Class of 1980 and 0.215 for the Class of 1992. It will come as no sur- prise that even when a full set of controls are included in the model, 1980 graduates who took both types of remedial courses were significantly less likely to attend college. These students also generally earned less. A year after graduating, males from the class of 1992 who had taken Table 4 The impact of remedial courses in English and math on college attendance and earnings of high school graduates a Females Males Coef. T Stat Coef. T Stat. 1992 high school graduates Hourly wage rate in 1993 20.009 0.71 20.008 0.69 Earnings in 1993 20.061 1.29 20.118 3.05 1980 high school graduates Earnings in 1981 20.041 1.00 20.026 0.64 Earnings in 1982 20.032 0.76 20.033 0.83 Earnings in 1983 20.047 1.17 20.040 1.03 Earnings in 1984 20.076 2.20 0.042 1.40 Earnings in 1985 20.085 2.75 0.016 0.59 Prob. of college attendance in 8182 20.041 2.93 20.063 4.01 Prob. of college attendance in 8283 20.043 2.79 20.038 2.17 Prob. of college attendance in 8384 20.054 3.56 20.038 2.23 Prob. of college attendance in 1985 20.010 0.65 20.026 1.38 1972 high school graduates Earnings in 1973 20.069 0.66 20.093 1.34 Earnings in 1974 20.163 2.15 20.077 1.41 Earnings in 1975 20.082 1.15 20.014 0.30 Earnings in 1976 20.037 0.62 20.110 2.76 Earnings in 1977 0.002 0.03 20.056 1.50 Earnings in 1978 0.054 0.94 20.086 2.42 Earnings in 1979 20.017 0.29 20.050 1.45 Prob. of attending college fall 72 20.014 1.05 20.029 2.62 Prob. of attending college fall 73 20.005 0.38 20.024 2.28 Prob. of attending college fall 74 0.001 0.10 0.003 0.04 Prob. of attending college fall 75 0.010 0.98 20.007 0.68 Prob. of attending college fall 76 20.008 0.94 20.010 1.12 Source: Analysis of NLS72, HSB and NELS88 data. a Coefficients of the earnings models have been divided by mean earnings for that year. Controls include past and present college attendance, reading and math test scores in 12th grade, grade point average, courses taken in high school, extra-curricular activities, work for pay during senior year, TV and homework hours, religion, reading for pleasure, attitudes, indicator for being handicapped, family demographics, marital and parental status at the end of 12th grade and region. p,0.01 on a two tail test; p,0.05 on a two tail test; p,0.10 on a two tail test. both types of basicremedial courses were earning a sig- nificant 12 less than those who had no basic courses. For females in the class of 1992 the estimate of the dif- ference is 6, a number that despite its large size is not significantly different from zero. For 1980 and 1972 graduates, taking remedial courses is negatively related to earnings in 20 of 24 comparisons possible. Five of the coefficients are significantly negative with a p value lying below 0.025 for a one tail test. While these results suggest that fewer basic and remedial courses may be a good thing, one should not take them literally as esti- mates of the effect of eliminating the option of taking remedial courses. Since 12th grade test scores, grades and attitude indexes are outcomes of earlier course selec- tion decisions, we may have attributed to these other variables effects that more properly were the result of course selection decisions. This would tend to bias the coefficients on this variable toward zero. On the other 344 J.H. Bishop et al. Economics of Education Review 19 2000 333–349 hand, stable unmeasured characteristics such as low ability or laziness may be the reason students are taking basic and remedial courses. Selection effects of this type would bias the coefficient in the negative direction away from zero. The findings presented in Sections 2 and 3 suggest that curriculum-based external exit exams like New York State’s Regents not only enhance student achievement, but also increase the earnings of graduating seniors even when measures of student achievement are held constant. So far, however, participation in Regents courses and exams has been voluntary. Requiring all students to take and pass Regents examinations in five core subjects to graduate from high school substantially raises the stakes. Regents examinations are considerably more challenging than the RCTs they will replace. Many students will find them very difficult. If they are to have a chance of pass- ing, schools will have to make the middle and high school curriculum more demanding and significantly increase the time and attention given to struggling stu- dents. 4. All-Regents high schools: how did they do it?