54 F.G. Mixon Jr, R.W. McKenzie Economics of Education Review 18 1999 51–58
The third variable, PhD, is a binary variable equal to one for institutions offering PhD programs of any kind,
and zero otherwise. These universities often receive higher levels of state funding on a per-student basis, and
may be policed by legislators, university trustees, fac- ulty, students, etc. more heavily than non PhD-granting
counterparts. It is also possible that higher per-student budgets are the result of greater monopoly power or rent-
seeking efforts by universitiescolleges.
3
Therefore, the sign of PhD is an empirical question. An additional
regressor, Roll, is equal to the size of the full-time stud- ent body in 1994 see America’s Best Colleges. We real-
ize that the size of institutions often changed over the history of each institution, however we argue that the
size differences among institutions in our sample was relatively undisturbed over the sample period. Because
larger institutions are more likely to have multiple layers of administration with more institutionalized decision-
making, thus reducing managerialpresidential responsi- bility for errors, it is likely that Roll will retain a positive
coefficient, ceteris paribus.
4
A price variable, AR, is average revenue for each institution in the sample for 1994. It is obtained by divid-
ing total tuition receipts by the size of the student body, and avoids tuition differences for in-state and out-of-
state students at certain institutions. We also make the same argument regarding relative changes versus absol-
ute changes for AR and any other independent variables that are measured with 1994 figures. We expect that AR
serves as a proxy for school quality which presidents have scope for control over and are rewarded for quality
maintenance, so that AR is expected to retain a positive sign, ceteris paribus. It is also likely that students who
pay more will reward more efficient management behavior.
FA denotes the percent of an institution’s student body that receives financial aid from any source for 1994. FA
represents, potentially, a source of managerial policing from the financial services sector, state governments, and
even the federal government through administration of financial aid from these various sources. These insti-
tutions may be more aware of the managerial practices of universities where FA is high, and may exert pressure
accordingly. However, government state and federal bureaucracies are political firms also, and may fail to
exert the type of influence that will promote efficiency.
3
Couch et al. 1992 show that a loophole in Alabamas Code of Ethics, which allows sitting state legislators to be on
the payroll of public colleges and universities, gives incentives for participating legislators to act as brokers by providing an
estimated 19 in additional public funding for every 1 received in salary.
4
Data on the number of administrators per student would be a better measure here, however we were unable to obtain this
information for much of our sample.
Therefore, the expectation regarding FA is ambiguous. Finally, ACT is included on the right-hand side, and
measures the average ACT score for incoming freshmen. ACT likely represents pressure from students — as with
the variable AR, good students national merit scholars and honors students will recognize and punish reward
inefficient managerial behavior. Therefore, we expect that ACT will retain a negative sign. All data for the
independent variables are measured in 1994 values and come from America’s Best Colleges.
5
4. Statistical results
The fully specified model was run in OLS along with several more parsimonious versions for comparison.
These results are included below in Table 2. Of the seven regressors included in model 1, three are significant at
the 95 level of confidence. The main regressor, Public, retains the expected positive sign and is significant at the
99 level of confidence in all versions of the equation. In fact the parameter estimate for Public ranges from
4.68 to 4.85, suggesting that the average tenure of presi- dents at public institutions is approximately 5 years
greater than their private counterparts, ceteris paribus. This significant finding supports the main premise of the
work of DeAlessi and Crain and Zardkoohi detailed above which suggests that managerial tenure under
government ownership is a source of incomeutility with significant present value, and public executives will
organize production and decision-making in a manner that rewards subordinates for loyalty while dissipating
executive responsibility for inefficiencies. These results suggest that university trustees and legislators should
consider compensation packages that provide incentives for managerial efficiency.
Of the other regressors in version 1, ACT is negative and significant and AR is positive and significant, which
are expected results. The main results are consistently found in other versions of Eq. 1, when some of the
regressors are deleted for comparison. The regression R- square values range from 19.1 to 21.2, as pointed
out at the bottom of Table 2. Version 1 of Table 2 is also employed to provide estimates for predicted average
managerial tenure across the sample of 73 institutions. These predictions are presented in Appendix A, and are
compared with the actual mean presidential tenure for each institution.
5
We realize that the use of such end-year 1994 data is a limitation of our test, because the factors influencing the tenure
of presidents have changed over time for each university in the sample. Although our data do not precisely capture these
changes, we do argue that stable relative values from school to school allow us to capture some of the important relationships.
55 F.G. Mixon Jr, R.W. McKenzie Economics of Education Review 18 1999 51–58
Table 2 Summary of OLS results. Dependent variable: tenure
1 2
3 4
intercept 13.88
13.79 13.63
13.78 2.54
2.54 3.35
3.41 Public
4.68 4.82
4.85 4.80
2.79 2.94
3.24 3.23
Urban 0.41
0.43 0.44
0.46 0.47
0.49 PhD
2 1.76
2 1.43
2 1.43
2 1.39
21.49 21.45
21.46 21.44
Roll 0.07E 2 3
0.51 AR
0.06E 2 2 0.06E 2 2
0.06E 2 2 0.06E 2 2
3.94 3.99
4.09 4.09
FA 0.002
2 0.001
0.07 20.05
ACT 2
0.43 2
0.40 2
0.40 2
0.39 22.01
21.95 22.04
22.01 nobs
73 73
73 73
F -statistic
2.50 2.91
3.54 4.07
R -square
0.212 0.209
0.209 0.191
Adj. R-square 0.127
0.137 0.150
0.144 Numbers in parentheses represent t-values for the OLS regression; [] represents significance at the 99[95] level of confi-
dence.
As a further test of the hypothesis, a new dependent variable, Pres was created which denotes the number
of presidents that each of the 73 institutions have had throughout their history. Because this variable does not
adjust for age, Found or the year in which the insti- tution was founded is included as a right-hand side vari-
able. This variable is expected to be negatively related to the new dependent variable Pres. Each of the other
regressors is included, and each is expected to retain the opposite
coefficient from Eq. 1 above. Also, because the dependent variable represents discrete values only,
this hypothesis is modeled with a maximum likelihood procedure tobit. The results are provided in Table 3.
The evidence provided in Table 3 provides further and perhaps stronger support for the hypothesis
developed by DeAlessi and Crain and Zardkoohi. Five of the eight regressors are significant at the 95 level
of confidence, and none retains an unexpected sign. The model suggests that public institutions have fewer presi-
dents over their lifespan, as expected, and that external institutions FA and students ACT possibly serve as
pressure groups on the managerial efficiency of univer- sities and colleges. In sum, the statistical evidence sup-
ports the view that job tenure is an important source of pecuniary and non-pecuniary incomeutility for presi-
dents managers of institutions of higher learning in the United States. From a public policy perspective, execu-
tive compensation schemes that reward cost savings, attracting quality students, and other amenities may work
Table 3 Empirical results of maximum likelihood tobit procedure.
Dependent variable: Pres Variable
Coefficient t
-ratio Intercept
14.97 6.86
Public 2
0.42 2
2.97 Urban
2 0.04
2 0.55
PhD 0.04
0.36 Roll
0.1E24 1.16
AR 2
0.4E24 2
3.05 FA
0.5E22 2.00
ACT 0.04
2.05 Found
2 0.01
2 6.78
Log-likelihood 2
18.54 Note: [] denotes significance at the 99[95] level of
confidence.
to promote efficiency. Current arguments for privatiz- ation or direct government subsidies to prospective stu-
dents in the form of vouchers also provide avenues to explore which also result in incentives that lead to
efficiency gains.
5. Concluding comments