Bayesian Inference in Public Administration Research: Substantive Differences from Somewhat Different Assumptions

Abstract

The purpose of this article is to point out that the standard statistical inference procedure in public administration is defective and should be replaced. The standard classicist approach to producing and reporting empirical findings is not appropriate for the type of data we use and does not report results in a useful manner for researchers and practitioners. The Bayesian inferential process is better suited for structuring scientific research into administrative questions due to overt assumptions, flexible parametric forms, systematic inclusion of prior knowledge, and rigorous sensitivity analysis. We begin with a theoretical discussion of inference procedures and Bayesian methods, then provide an empirical example from a recently published, well-known public administration work on education public policy.

Notes

19. Geman, S.; Geman, D. Stochastic Relaxation, Gibbs Distributions and the Bayesian Restoration of Images. IEEE Transactions on Pattern Analysis and Machine Intelligence 1984, 6, 721–741; Smith, A.F.M.; Roberts, G.O. Bayesian Computation Via the Gibbs Sampler and Related Markov Chain Monte Carlo Methods (with discussion). Journal of the Royal Statistical Society, Series B 1993, 55, 3–24

20. For a comprehensive discussions of the theory and practice of Gibbs sampling and MCMC in general see: Carlin, B.P.; Louis, T.A. Bayes and Empirical Bayes Methods for Data Analysis, 2nd Ed.; Chapman amp; Hall: New York, 2000; Gamerman, D. Markov Chain Monte Carlo; Chapman amp; Hall; New York, 1997; Gelfand, A.E.; Smith, A.F.M. Sampling-based Approaches to Calculating Marginal Densities. Journal of the American Statistical Association 1990, 85, 389–409; Gelman, A.; Rubin, D.B. Inference from iterative simulation Using Multiple Sequences. Statistical Science 1992, 7, 457–511; Gelman, A.; Carlin, J.B.; Stern, H.S.; Rubin, D.B. Bayesian Data Analysis; Chapman amp; Hall: New York, 1995; Geweke, J. Bayesian Inference in Econometric Models Using Monte Carlo Integration. Econometrica 1989 57, 1317–1339; Gill, J. Bayesian Methods: A Social and Behavioral Sciences Approach; Chapman amp; Hall: New York, 2002; Tanner, M.A. Tools for Statistic Inference: Methods for the Exploration of Posterior Distributions and Likelihood Functions; Springer-Verlag: New York, 1996.

39. Ibid.

40. Ibid.

41. Beck, N.; Katz, J.N. Random Coefficient Models for Time-Series-Cross-Section Data: The 2001 Version. Available at the Political Methodology Working Paper Archive: http://web.polmeth.ufl.edu/,2001

42. Meier, K.J.; Polinard, J.L.; Wrinkle, R. formance: Opcit, p. 595.

44. Hanushek, Eric A. The Economics of Schooling: Production and Efficiency in Public Schools. Journal of Economic Literature 1986, 24, 1141–1177; Monk, D. Education Productivity Research: An Update and assessment of its Role in Education Finance Reform. Educational Evaluation and Policy Analysis 1992, 14 (4), 307–332; Ferguson, R.F. Paying for Public Education: New Evidence on How and Why Money Matters. Haryard Journal on Legislation 1991, 28, 465–498; Ferguson, R.F.; Ladd, H.F. How and Why Money Matters: An Analysis of Alabama Schools. In Holding Schools Accountable: Performance-based Reform in Education; Ladd, H. F., Ed.; Brookings Institution Press: Washington, D.C., 1996.

45. Aigner, D.J.; Chu, S.F. On Estimating the Industry Production Function. American Economic Review 1968, 58, 826–837; Comanor, W.S.; Leibenstein, H. Allocative Efficiency, X-Efficiency, and the Measure of Welfare Losses. Economica 1969, 26, 304–309; Timmer, C.P. Using Probabilistic Frontier Production Function to Measure Technical Efficiency. Journal of Political Economy 1971, 79 (4), 776–794.

49. Lee, P.M. Bayesian Statistics: An Introduction; Oxford University Press: New York, 1989; Pollard, W.E. Bayesian Statistics for Evaluation Research: An Introduction; Sage: Beverly Hills, 1986.

50. For theoretical details review: Rubin, D. Multiple Imputation for Nonresponse in Surveys; John Wiley amp; Sons: New York, 1987; Little, Roderick J.A. and Rubin, D.B. On Jointly Estimating Parameters and Missing Data by Maximizing the Complete-Data Likelihood. The American Statistician 1983, 37, 218–220.

55. The term β was used for ease of computation and reference. Each numbered β represents the prior information assigned to each explanatory variable. For reference: beta[0]=constant; beta[1]=Low Income Students; beta[2]=Teacher Salaries; beta[3]=Teacher Experience; beta[4]=Gifted Classes; beta[5]=Class Size; beta[6]=Percent State Funding; beta[7]= Funding Per Student; beta[8]=Lag of Student Pass Rate; beta[9]=Lag of Beaucrats; betas[10-15] represent the years 1993–1997.

59. Meier, K.J.; Polinard, J.L.; Wrinkle, R. Opcit, p. 593.

60. The prior for the interaction variable was operationalized in the model as beta [16].

71. Meier, K.J.; Polinard, J.L.; Wrinkle, R. Opcit, 590–602. See also: Downs, A. Inside Bureaucracy; Little and Brown: Boston, 1967.

80. Western, B.; Jackman, S. Opcit.