The Money Myth: School Resources, Outcomes, and Equity

Appendix A

Appendix A Technical Issues andVariable Defınitions THESE RESULTS are based on the restricted-use sample of NELS88. Three complications arise in estimating equations with NELS88 data. One is the problem of missing data for individual questions. One potential solution— deleting every observation with missing data—would both bias the sample and result in a much smaller sample. In these results, I give each individual with missing data for a particular variable the mean value from the rest of the sample and construct a dummy variable equal to one for those with missing data; this approach exhausts the information available. An alternative , multiple imputation, involves estimating equations for each of the variables for which there are missing data and using these estimated equations to interpolate values (Rubin 1987), but in addition to this method’s complexity , there is no causal logic underlying the estimated equations. A second complication is that NELS88, like every longitudinal data set, lost students over time. NELS88 began with 24,599 individuals in eighth grade;that sample dwindled to 16,489 with complete data for the first three waves and 13,120 for those with the first four waves of data, including students in both public and private schools.The results in this volume, for outcomes as of twelfth grade, start from the second of these samples, though missing data for various dependent variables means that the sample sizes for particular specifications vary from this (see the numbers in table B.2). In addition , the dependent variables for high school graduation and for entry into two- and four-year colleges were drawn from transcript data and the third follow-up, respectively, so these sample sizes are smaller still. Because attrition was nonrandom, the data set includes weights to compensate for attrition so that the sample when weighted can mimic a nationally representative sample. Most results for second follow-up (F2) outcomes in this paper are weighted by F2PNLWT, and for results based on third follow-up (F3) outcomes with F3PNLWT.The use of weights makes a substantial difference to parameter estimates, so software without the capacity to incorporate weights was rejected. In addition, NELS collected data with a two-stage sampling procedure: schools were randomly sampled,and then students within schools were randomly sampled. Because the students within particular schools are not as varied as students in the entire population—since schools are segregated by family background, including income, parental education, and race—the variation in such a sample is smaller than in the population, and the standard errors of estimated parameters are also smaller than those from a random sample.There are various approaches to correcting standard errors; the results presented here have been estimated by STATA, which uses aTaylor series approach to calculating standard errors. In the results presented here, all variables are taken from the second follow -up, senior year, unless noted. P refers to the parent questionnaire, S to the student questionnaire,A to the administrator questionnaire, andT to the teacher questionnaire; these provide information about sources of information . “Adj.” refers to adjustment for cross-sectional cost differences by Chambers’s (1998) measures plus the consumer price index (CPI) for timeseries variation, since the CPI is as effective as more complicated indices of inflation (Chambers 1997). DependentVariables:Schooling Outcomes MATHTS Math test score, scaled F22XMIRR SCITS Science test score, scaled F22XSIRR READTS Reading test score, scaled F22XRIRR HISTTS History test score, scaled F22XHIRR CLSRNK Class rank (percentile) F2RRANK HEDASP High educational aspirations S43 HOCASP High occupational aspirations S64 CONTED Plans to continue education past high school S56 TOTCRED Total credits earned F2RNWB2A ACPRO Completed standard academic program F2RNWB3A DIPLOM High school diploma (transcriptreported ) F2RTROUT ENR4YR Enrolled in four-year college (third follow-up) PSEFIRTY ENT2YR Enrolled in two-year college(third follow-up) PSEFIRTY 290 The Money Myth IndependentVariables School Resources: Simple Resources P/TRTIO (—) Pupil-teacher ratio A1,A29 LOWTCHSAL Lowest teacher salary A37, adj. HITCHSAL Highest teacher salary A37, adj. Compound Resources TCHEXP Average years taught in secondary school T4.4b INFIELDTCH Teaching in field T4.9,T4.10 PLANTIME Teacher planning time A38 STDEV Staff development time T4.17 GENED Student in the general track S12a VOCED Student in the vocational track S23a REMED Student in remedial courses S13a, b, f, h Complex Resources TIMEUSE Percentage of time on order, tests, administration T2.12d, e, f CONVTCH Conventional teaching approach T2.13a, c, e, h INNOVTCH Innovative...