Page 1

The Dog ATE my Economics Homework!

Estimates of the Average Effect of Treating

Hawaii’s Public High School Students with Economics*

Kimberly Burnett

University of Hawaii Economic Research Organization

University of Hawaii-Manoa

and

Sumner La Croix

Department of Economics

University of Hawaii-Manoa

Working Paper No. 10-1

January 22, 2010

Abstract

Hawaii is one of 27 states that do not require testing of public high school students

regarding their understanding of economics. We report results for the first economics test

administered to a large sample of students in Hawaii public high schools during the

Spring 2004 semester. Our analysis focuses on evaluating the impact of a semester-long

course in economics on student scores on a 20-question, multiple-choice economics test.

We specify and estimate a regression analysis of exam scores that controls for other

factors that could influence student performance on the exam. While student scores on

the economics exam are relatively low, completion of an economics course and

participation in a stock market simulation game each add about one point to student

scores.

Keywords: economic education, high school economics, stock market simulation

JEL Codes: A20, A21, I21

* We thank Sang-Hyop Lee, Tim Halliday, Michael Kimmit, and participants in a seminar at the University of Hawaii for useful

comments. We thank Kristine Castagnero, Director of the Hawaii Council for Economic Education, for her early comments on this

research and support of this project. Gail Tamaribuchi, former Director of the University of Hawaii Center for Economic Education,

was instrumental in identifying high school teachers who were willing to administer the exam to their classes. We thank the public

high school teachers and students who took part in this project and the State of Hawaii Department of Education for allowing us to

administer a test to students in public high school classes. Finally, the views expressed in this paper are solely those of the authors

and do not necessarily reflect those of the University of Hawaii Dept. of Economics, the University of Hawaii Economic Research

Organization, or the Hawaii State Department of Education.

Page 2

1

I. Introduction

Over the last 50 years, U.S. public and private high schools have increasingly

incorporated economics courses into their curriculums. For the 2004-2005 academic

year, the High School Transcript Study (2007) reported that 45.8 percent of high school

graduates had completed an economics course.1 Standardized tests of nationwide

samples of high school seniors have been conducted since the early 1970s to measure

understanding of key economic concepts and ability to apply them to particular problems.

While average student scores have been relatively low, completion of an economics

course has substantially raised student performance (Walsted and Rebeck 2000, 2001).

In 2006, the National Center of Education Statistics conducted its first assessment of

economics knowledge for the National Assessment of Educational Progress (NAEP) by

testing 11,500 twelfth-grade students from 590 public and private high schools. Walstad

and Buckles (2008) report cross tabulations from the 2006 NAEP assessment that suggest

important demographic, socioeconomic, instructional, and aptitude-related determinants

of exam performance.2

Twenty-three states require high school students to be tested on their knowledge

of economics.3 There appears to be little systematic reporting of results from the states

requiring testing, while statewide tests in economics are almost never conducted in the

states without required testing. Hawaii is one of the states in which economics is well

established in the public high school curriculum and which has never administered a

standardized economics exam to its public high schools students or to students who

complete an economics class. The lack of testing is somewhat surprising, as over 90

percent of Hawaii public high schools typically offer a full-semester economics course

and 27 percent of high school seniors completed a semester course in economics in both

the 2004-2005 and 2006-2007 academic years.4 Test results for students within a

particular state and at particular high schools could, however, help administrators and

policymakers to assess the overall understanding of economics by high school graduates

and evaluate the impact of an economics course on a student’s economic knowledge.

We report results for the first economics test administered to a large sample of

students in Hawaii public high schools during the Spring 2004 semester. Our analysis

1 The National Center for Education Statistics (2007) reported that 66 percent of high school seniors in

2006 had taken either a general economics course or an advanced placement economics course.

2 The NAEP does not release state-by-state breakdowns due to the small number of students taking the

exam in some states.

3 Grimes et al. (2008) suggest that the increases in economics exam scores are smaller for students taking a

mandated economics course than for students taking the course as an elective.

4 Hawaii public high school students are required to complete four credits (8 semesters) of social studies

courses. Since 3 credits are taken up by required social studies courses, students complete their one

elective social studies credit with two semester-length courses chosen from social studies electives offered

by their high schools, e.g. psychology, geography, economics, consumer education, or European history.

Page 3

2

focuses on evaluating the impact of a semester-long course in economics on student

scores on a 20-question, multiple-choice economics test. We specify and estimate a

regression analysis of exam scores that controls for other factors that could influence

student performance on the exam. As we discuss below, the State of Hawaii placed

severe restrictions on the quantity and type of questions that we could ask high school

students regarding their personal and family characteristics. As a consequence, some

variables that affect student’s selection of an economics elective and their performance

on our test could not be included in the regression analysis. Despite these limitations, we

are able to draw some conclusions from our regression results regarding the impact of an

economics course on student performance on our economics test.

II. Exam Methodology

We analyze results from a 20-question multiple-choice test on basic economics

administered to over 500 students enrolled in 19 public high schools in Hawaii during the

Spring 2004 semester. The 19 high schools were drawn from Hawaii’s 38 public high

schools based on the willingness of an individual teacher at each high school to

administer the exam to the teacher’s class. The exam was administered at high schools in

each of Oahu’s four school districts and from the neighbor islands of Hawaii, Maui,

Molokai, and Kauai. The exams were conducted in a high-school class but not in an

economics or a consumer education class. The exam did not count towards a student’s

grade in the class. No compensation was paid to students either for completion of the

exam or performance on the exam. Each class had some students who had completed an

economics course and some who had not.5

The 20-question multiple-choice exam was designed by the National Council for

Economic Education, an umbrella organization of state council of education. The

questions cover such topics as exchange, supply and demand, price controls, inflation,

national income accounting, and international trade. Each question had three possible

answers.

We asked the Hawaii State Department of Education (“Hawaii DOE”) for

permission to ask students 20 questions regarding their personal and family backgrounds.

The Hawaii DOE did not allow us to ask any questions pertaining to their family (e.g.

number of siblings, lives with both parents, each parent’s occupation, high school GPA,

ethnicity, and whether English was their native language or spoken at home. In fact, the

Hawaii DOE allowed us to ask just four questions pertaining to their personal

characteristics: (1) gender; (2) plans to attend a two-year college; (3) plans to attend a

four-year college; and (4) plans to take more economics after high school. The Hawaii

DOE denied permission to ask students for their high school GPA, ethnicity, and whether

English was their native language.6

5 In one high school, we learned after the test had been conducted that all students who took the economics

exam had completed an economics class. We dropped this high school from our sample.

6 The Hawaii Dept. of Education routinely allows such information to be provided to national testing

agencies.

Page 4

3

We also asked students whether they had completed (1) a semester-length course

in economics; (2) a partial course in economics; (3) a consumer education course; and (4)

the Hawaii Stock Market Simulation. We asked specifically about these two courses and

classroom exercises because each may teach some concepts and provide some

information that could improve performance on an economics exam.7 After reviewing

student responses for partial courses in economics, we found that the only students who

responded that they had completed a partial course in economics were students who had

also completed a semester-length course in economics. We conclude that students were

confused by this question and have not included this variable in our regression analyses.

The Hawaii Stock Market Simulation (SMS), administered by the Hawaii Council

on Economic Education, is an interactive educational program that teaches high school

students about U.S. securities markets. Student teams invest a hypothetical $100,000 in

stocks, bonds, and mutual funds over a 10-week period. Team winners are publicized in

the Hawaii media.8 The intent of the program is to encourage students to think about real

life investment and financial issues such as setting investment goals, their tolerance to

risk, and the trade-offs that must be considered as they determine how to save for their

future.

IV. Sample Properties and Summary Statistics

We examined all test forms and identified 25 with particularly low scores that

were likely due to a lack of effort on the exam. A large number of questions were left

blank on some exams; others exhibited a pattern of answers inconsistent with efforts to

honestly answer each question. Three high schools each had four students with

questionable exams. Only 20 percent of these students had taken an economics course,

and only 8 percent (2 students) had participated in a stock market simulation. These

students were also more likely to be male (56 percent) and were less likely to have plans

to attend a four-year college. We decided to drop these exams from our regression

analysis. After dropping these exams, we find that student scores on the economics test

are normally distributed with no discernable heaping at the tails.

7 Our review of consumer education courses in Hawaii shows them to be focused on such issues such as

consumer budgeting, checkbook balancing, credit management, consumer protection, and personal saving

and investment.

8 Two 10-week sessions are offered each school year. Students participate in divisions based on their

investor profile: Income Growth, Growth, and Aggressive Growth. Each division requires that students

maintain a certain asset allocation of stocks and bonds. For example, a student participating in the Income

Growth Division would build a portfolio that contains 50 percent common stock and/or equity mutual funds

and 50 percent bonds and/or bond funds. At the end of the simulation, the Hawaii Council on Economic

Education reviews each portfolio (based on the division in which the team is participating) and compares

its performance to an investor profile benchmark (i.e. Growth & Income) representing the proper asset

allocation of the 10 year treasury bond and the S&P500.

Page 5

4

We compute summary statistics for (1) the entire sample of students; (2) students

without a full economics course; and (3) students with a full economics course. Thirty-

seven percent of students had completed a semester-length economics; 24 percent had

participated in the Stock Market Simulation; and 19 percent had completed a semester-

length class in consumer education. The summary statistics show that the students who

selected an economics class were somewhat different than the students who did not.

They were more likely to be male (53 percent vs. 42 percent), more likely to plan to

attend a four-year college (62 percent vs. 51 percent), less likely to attend a two-year

college (18 percent v. 26 percent) and more likely to have participated in the Stock

Market Simulation (42 percent vs. 14 percent). The difference in characteristics between

the two groups of students clearly indicates that students are self-selecting into the class,

and that we cannot view an economics class as a randomly assigned treatment.9

IV. Econometric Methodology

Our goal in estimating regressions on student test scores is to isolate the effect of

an economics course and a stock market simulation on student understanding of

economics. Since students are clustered within specific high schools (“strata” in the

econometrics literature), our econometric analysis uses a specific form of this more

general strata regression:

(1)

SCOREis= α + β1ECON _COURSEis+ β2STOCK _ MKT_SIMis+γICis+κSCs+qs+eis

where SCOREis is the number of questions correct on the economics test by student i at

school s,

αs is a school-specific intercept, and ECON_COURSEis and

STOCK_MKT_SIMis are the two treatment variables,

control variables, SCs is a vector of strata-specific control variables, qs is an unobserved

stratum effect, and eis is the error term for student i at school s.

Wooldridge (p. 133) noted that “the presence of the unobservable qs induces

correlation in the composite error term

µis= qs+εiswithin each stratum.” Estimating

individual and stratum effects in one regression leads to unbiased estimators for β, γ, and

κ but consistency and asymptotic normality cannot be demonstrated. Moulton (1992)

demonstrated that within-group correlation leads to upward biased standard errors for

and

ˆ γ . However, since our analysis focuses on estimated coefficients for individual

rather than strata characteristics, adding a set of strata dummies to the regression provides

a simple solution to the error correlation problem.10 This yields a more specific form of

strata regression:

9 Peterson (1992) presented evidence that failing to account for the self-selection bias will understate the

potential gain attributable to a course in economics.

10 Once strata dummies are added to the regression, adding additional variables covering strata

characteristics to the regression do not improve the efficiency of estimates for variables covering student

characteristics.

€

€

€

ICisis a vector of student-specific

€

€

ˆ β

€

Page 6

5

(2)

where

€

SCOREis= αs+ β1ECON _CLASSis+ β2STOCK _SIMis+γICis+eis

€

αs is a stratum-specific intercept.

The dependent variable in all regression specifications—the student’s score on the

multiple choice exam, SCOREis—is count data, with a minimum value of 0 and a

maximum of 20. Estimated coefficients from a regression with a count-data dependent

variable often have superior properties when a Poisson or a negative binomial estimator

is used rather than OLS (Wooldridge, ch. 19). Standard goodness of fit tests indicated

that the negative binomial model was more appropriate than the Poisson model. Thus,

we estimate all regression specifications using both OLS and negative binomial

estimators.

The severe restrictions imposed by the Hawaii Department of Education on the

amount of personal data that could be gathered from each student has the potential to bias

estimated coefficients on both treatment variables due to the usual problems stemming

from omitted control variables. If data were available, we would have included high

school GPA, ethnicity, and whether English was their native language as control

variables in each regression specification. The small number of student characteristic

variables also limits our ability to control for selection problems with the two non-

randomly assigned treatment variables.

Each regression includes three controls for student characteristics—gender, plans

to attend a two-year college, and plans to attend a four-year college. Our basic regression

specification follows:

SCOREis= αs+ β1ECON _COURSEis+ β2STOCK _ MKT_SIMis+γ1GENDERis

+γ2TWO_YR_COLLEGEis+γ3FOUR_YR_COLLEGEis+eis

Finally, the students who took an economics class or participated in a stock

market simulation were not randomly assigned to these treatments but rather self-selected

into them. If more information on student (and their family’s) characteristics were

available, it might be possible for to use instrumental variable or matching estimation

techniques to address this problem. In lieu of this information, we construct a second

data set that only includes groups of students with identical characteristics (gender, post-

secondary education plans, and high school) and with at least one student with a different

value for the ECON_COURSE treatment variable. All other observations are dropped.

By dropping non-comparable observations from the data set, there is the potential for

selection bias to be reduced. We estimate the following regression using OLS and

negative binomial estimators with the common support data set:

(4)

SCOREis= αg+ β1ECON _COURSEig+ β2STOCK_ MKT_SIMig+eig

where αg is a group-specific intercept. Results from regressions with the common

support data set need to be considered cautiously, as a matching analysis will only

perform well if selection is solely on observables (Imbens and Wooldridge 2009). This

(3)

€

€

Page 7

6

condition is clearly not met in our analysis, as we have strong priors that other variables

(e.g. family income, student GPA) are likely to affect student selection of an economics

class or a class with a stock market simulation.

Our econometric analysis begins (Table 2) with OLS estimates of specification

(3) and considers whether results are affected by adding an additional treatment variable

(CONS_ED) and an additional control variable (PLAN_ ECON). Next, we run the same

regression specifications with a negative binomial estimator because the dependent

variable (SCOREis) is count data (Table 3). Finally, we repeat both the OLS and negative

binomial estimations using the matched data set described above (Table 4).

V. Results

Our empirical analysis of student test scores begins with an examination of results

from four different regression specifications estimated using ordinary least squares. In

each of the four specifications (Table 2, columns 1-4), control variables performed as

expected. Estimated coefficients on GENDER range between .73 and .75 and are

statistically significant at the five percent level in all specifications. These results are not

surprising, as they mirror those found in earlier national studies of performance of high

students on economics tests, in particular the recent NAEP assessment in 2006.

Estimated coefficients on TWO_YR_COLLEGE are positive, ranging from .42 to .44,

but are not statistically significant at the ten percent level in all specifications. Estimated

coefficients on FOUR_YR_COLLEGE are also positive, ranging from 1.33 to 1.35, and

are statistically significant at the five percent level. Both variables are measured against

the baseline of the group of students who have no plans for further education beyond high

school. These results support the finding by Walstad and Buckles (2008) that students

with higher post-high school aspirations are likely more academically inclined than

otherwise, and therefore perform better on the exam.

All OLS specifications were run with dummy variables for each student’s high

school to control for stratum effects.11 None of the estimated coefficients are statistically

significant at the ten percent level. This is a somewhat surprising finding, as our priors

indicate that the quality of instruction varied substantially across high schools, a factor

that should have produced a positive impact on test scores at these schools.

Two more OLS specifications (Table 2, columns 2 and 4) were run using an

additional control variable, PLAN_ ECON. Estimated coefficients on PLAN_ ECON

range from .06 to .14, but are not statistically significant at the ten percent level.

Since the control variables generally performed as expected (or were statistically

insignificant), we turn our focus to estimates of the effects of the two treatment variables,

ECON_COURSE and STOCK_MKT_SIM. Estimated coefficients on ECON_COURSE

range from 1.02 to 1.09, and are statistically significant at the five percent level in all four

specifications. We note that the mean of the four conditional estimates of the effect of an

11 One school dummy was omitted to prevent perfect collinearity.

Page 8

7

economics course on a student’s score (1.05) is less than the difference in unconditional

means (1.50). Estimated coefficients on the second treatment variable,

STOCK_MKT_SIM ranged from .96 to 1.04, and are statistically significant at the five

percent level in all four specifications. Together, the two treatments raised student test

scores by just over two points. These results support findings by Walstad and Buckles

(2008) that participation in a stock market simulation reinforces economic concepts and

therefore improves performance on the test.

Estimated coefficients on a third treatment administered to some students,

CONS_ED, are negative (Table 2, columns 3 and 4). This is somewhat surprising as one

might expect some of the knowledge acquired in a consumer education course to be

useful on a test of economics knowledge.12 The estimates are, however, not statistically

significant at the ten percent level.

Student test scores on the economics test are count data, and OLS estimates with

count data could be biased. We experimented with a Poisson regression model, but ran

into issues with overdispersion, i.e., greater variance than might be expected in this type

of distribution, resulting in failures of standard goodness-of-fit tests.13 We followed the

Poisson model with a negative binomial regression, as it is often more appropriate when

there is overdispersion. A likelihood ratio test conducted on the negative binomial

regression results reinforces our earlier finding that the assumption of a Poisson

distribution is inappropriate for our data set. Table 3 reports marginal effects of

treatment and control variables on student score for the same four specifications used in

the OLS estimates. Inspection quickly reveals that results from the negative binomial

regressions are virtually identical to those from the OLS regressions.

We also run OLS and negative binomial regressions on two specifications (with

and without CONS_ED) with our matched data set (Table 4). Estimated coefficients for

ECON_COURSE range from 1.01 to 1.08 and are statistically significant at the five

percent level. These results are just about the same as those obtained from earlier OLS

and negative binomial regressions using the full data set. Estimated coefficients for

STOCK_MKT_SIM range from 1.03 to 1.11 and are statistically significant at the five

percent level. These results are slightly (.08 to .10) higher than those obtained from

earlier OLS and negative binomial regressions using the full data set.

VI. Conclusion

Our empirical results for the effects of the stock market game and completion of

an economics course are robust across two different estimation methodologies (OLS and

negative binomial regression), estimation with matched and unmatched data, and

12 Walstad and Buckles (2008) find a similar result regarding the negative relationship between

participating in Junior Achievement and test scores.

13 A significant (p<0.05) test statistic from the goodness-of-fit test indicates that the Poisson model is

inappropriate. The large value for chi-square in our goodness-of-fit test was another indicator that the

Poisson distribution was not an adequate functional form.

Page 9

8

inclusion of a third treatment variable. In sum, they indicate that the Hawaii public high

school students in our sample who completed an economics course scored about one

point higher on the twenty-question exam and that students who participated in the stock

market simulation also scored about one point higher on the exam.

Participation in the stock market game produced almost as much improvement on

the student test as completion of an economics course. This is somewhat surprising, as

the amount of student and teacher time devoted to this activity is far less than the

semester-long economics course.

How do our results inform us with respect to the impact of a course in economics

on a Hawaii public high school student? On the one hand, scores of Hawaii public school

students on our economics test mirror nationwide results on standardized economics

tests: they are not particularly high. Students who completed an economics course still

averaged just 59 percent on the exam. On the other hand, Walstad (2001) suggests that

the best opportunity for improving the economic understanding of youth occurs in high

school, and our empirical results show non-trivial improvement, as the course in

economics added about five percent of the total points possible to student scores. While

the effect is not large, it nonetheless shows that foundations are present in the Hawaii

public high school curriculum to improve student understanding further. Finally, our

analysis emphasizes contemporaneous effects, but there could also be (unmeasured)

effects on future performance, as Myatt and Waddell (1990) found that completion of a

high school economics course enhances performance in university-level economics.

Page 10

9

References

Burnett, Kimberly, and Sumner La Croix (2009). “Economic Education’s Roller Coaster

Ride in Hawaii, 1965-2006.” Dept. of Economics University of Hawaii Working

Paper No. 09-1.

Grimes, Paul, Meghan Millea and M. Kathleen Thomas (2008). “District Level Mandates

and High School Students’ Understanding of Economics.” Journal of Economic

and Economic Education Research, May, 1 2008.

Imbens, Guido W., and Jeffrey M. Wooldridge (2009). “Recent Developments in the

Econometrics of Program Evaluation,” Journal of Economic Literature, 47(1)

March: 5-86.

Moulton, Brent R. (1992). “An Illustration of a Pitfall in Estimating the Effects of

Aggregate Variables on Micro Units,” Review of Economics and Statistics, 72(2)

May: 334-338.

Myatt, Anthony and Charles Waddell (1990). “An Approach to Testing the Effectiveness

of the Teaching and Learning of Economics in High School.” Journal of

Economic Education 21(3) Summer: 355-363.

National Center for Education Statistics (2005). America’s High School Graduates:

Results from the 2005 NAEP High School Transcript Study. Washington, D.C.:

Dept. of Education. Available online at http://nces.ed.gov/pubsearch/

pubsinfo.asp?pubid=2007467. Last accessed on 17 November 2008.

National Center for Education Statistics (2001). The High School Transcript Study: A

Decade of Change in Curricula and Achievement, 1990-2000. Washington, D.C.:

Dept. of Education. Available online at http://nces.ed.gov/pubsearch/

pubsinfo.asp?pubid=2004455 Last accessed on 17 November 2008.

National Center for Education Statistics (1998). The 1998 High School Transcript Study

Tabulations: Comparative Data on Credits Earned and Demographics for1998,

1994, 1990, 1987, and 1982 High School Graduates. Washington, D.C.: Dept. of

Education. Available online at http://nces.ed.gov/pubsearch/pubsinfo.asp?pubid

=2001498. Last accessed on 17 November 2008.

National Center for Education Statistics (2007). The Nation’s Report Card: Economics

2006. NCES 2007-475. Washington, D.C.: U.S. Dept. of Education.

National Council on Economic Education (2008). Survey of the States: Economic and

Personal Finance Education in our Nation’s Schools in 2007: A Report Card.

New York: National Council on Economic Education.

Page 11

10

Peterson, Norris A. (1992). “The High School Economic Course and its Impact on

Economic Knowledge.” Journal of Economic Education 23 Winter: 5-14.

Snyder, Thomas D and Charlene Hoffman. (2003) Digest of Education Statistics 2002.

National Center for Education Statistics, Report number NCES 2006-060.

Walstad, William B. (1992). “Economics Instruction in High School,” Journal of

Economic Literature 30(4) December: 2019-2051.

Walstad, William B. (2001). “Economic Education in U.S. High Schools,” Journal of

Economic Perspectives 15(3) Summer: 195-210.

Walstad, William and Ken Rebeck. (2001). “Assessing the Economic Understanding of

US High School Students,” American Economic Review 91(2): 452-457.

Walstad, William B., and Ken Rebeck (2000). “The Status of Economics in the High

School Curriculum,” Journal of Economic Education 31(1) Winter: 95-101.

Walstad, William B., and Stephen Buckles (2008). “The National Assessment of

Educational Progress in Economics: Findings for General Economics,” American

Economic Review 98(2) May: 541-546.

Wooldridge, Jeffrey M. (2004). Econometric Analysis of Cross-Section and Panel Data.

Cambridge: MIT Press.

Page 12

11

Table 1: Summary Statistics: Student Test Scores and Characteristics

A. All Students Taking the Exam

Unmatched Sample

Variable Mean

-----------------------------------------------------------------------------------------------------------

SCORE 10.88

ECON_COURSE .37

PART-ECON-COURSE .20

STOCK_MKT_SIM .24

CONS_ED_COURSE .19

GENDER .46

TWO-YR_COLL .23

FOUR-YR_COLL .55

PLAN_MORE_ECON .20

Observations 468

B. Students Taking Exam with Economics Course

Unmatched Sample

SCORE 11.83

ECON_COURSE 1.00

PART_ ECON .53

STOCK_MKT_SIM .42

COMBINATION .47

GENDER .53

TWO_YR_COLLEGE .18

FOUR_YR_COLLEGE .62

PLAN_ ECON .22

Observations 172

C. Students Taking Exam without Economics Course

Unmatched Sample

SCORE 10.33

ECON_COURSE .00

PART_ ECON .00

STOCK_MKT_SIM .14

COMBINATION .04

GENDER .42

TWO_YR_COLLEGE .26

FOUR_YR_COLLEGE .51

PLAN_ ECON .19

Observations 296

Matched Sample

Mean S.D. S.D.

3.67

.48

.40 .23

.43

.40

.50

.42

.50

.40

10.95

.42

3.77

.49

.42

.42

.38

.50

.39

.49

.39

.23

.17

.44

.19

.60

.19

305

Matched Sample

11.69

1.00

.32

.36

.53

.18

.58

.20

129

3.69

.00

.50 .53

.50

.50

.50

.39

.49

.42

3.81

.00

.50

.47

.48

.50

.38

.50

.40

Matched Sample

10.41

.00

3.55

.00

.00 .00

.35

.19

.49

.44

.50

.39

3.65

.00

.00

.37

.18

.49

.40

.49

.39

.16

.03

.38

.20

.62

.18

176

Page 13

12

Table 2: OLS Regressions on Student Test Scores:

Full Sample with School Dummy Variables

Variable

(1) (2) (3) (4)

ECON_COURSE

STOCK_MKT_SIM

CONS_ED

GENDER

TWO_YR_COLLEGE

FOUR_YR_COLLEGE

PLAN_ ECON

School Dummies

Adj. R2

Observations

F-Statistic

1.09**

(.44)

1.09**

(.44)

1.03**

(.44)

1.02**

(.44)

.96**

(.43)

.96**

(.43)

1.04**

(.44)

1.04**

(.44)

-.59

(.46)

-.61

(.46)

.73**

(.33)

.73**

(.33)

.75**

(.33)

.75**

(.33)

.43

(.49)

.42

(.49)

.44

(.49)

.43

(.49)

1.35**

(.42)

1.34**

(.43)

1.36**

(.42)

1.33**

(.43)

.06

(.42)

.14

(.42)

yes yes yes yes

.11

468

3.67

.11

468

3.50

.11

468

3.59

.11

468

3.43

Page 14

13

Table 3: Negative Binomial Regressions for Student Test Scores:

Full Sample with School Dummy Variables

Variable

(1) (2) (3) (4)

ECON_COURSE

STOCK_MKT_SIM

CONS_ED

GENDER

TWO_YR_COLLEGE

FOUR_YR_COLLEGE

PLAN_ ECON

School Dummies

Observations

Pseudo R2

Log Likelihood

LR χ2(23)

Note:

1.11**

(.43)

1.11**

(.43)

1.05**

(.44)

1.04**

(.44)

.94**

(.43)

.93**

(.43)

1.03**

(.44)

1.02**

(.44)

-.60

(.45)

-.62

(.45)

.72**

(.32)

.72**

(.32)

.74**

(.32)

.74**

(.32)

.46

(.51)

.45

(.51)

.47

(.51)

.46

(.51)

1.38**

(.42)

1.36**

(.43)

1.39**

(.42)

1.35**

(.42)

.08

(.41)

.16

(.41)

yes yes yes yes

468

.03

1241.3

76.94

468

.03

1241.2

76.98

468

.03

1240.4

76.73

468

.03

1240.3

78.88

Page 15

14

Table 4: OLS and Negative Binomial Regressions for Student Test Scores:

Matched Sample

Variable

OLS OLS NB NB

ECON_COURSE

STOCK_MKT_SIM

CONS_ED

School Dummies

Match Dummies

Observations

Adj. R2

F-Statistic

Pseudo R2

Log Likelihood

LR χ2(23)

Note: ** indicates statistical significance at the five percent level.

1.08**

(.47)

1.02**

(.47)

1.08**

(.44)

1.01**

(.44)

1.05**

(.54)

1.11**

(.54)

1.03**

(.51)

1.11**

(.51)

-.76

(.58)

-.77

(.54)

no

yes

no

yes

no

yes

no

yes

305

.09

1.80

305

.11

3.50

305

305

-807.06 -806.04

68.43

.04 .04

70.47