Money Only Matters if You Want It To? Exposing the Normative Implications of Empirical Research

• Author: J. Scott Granberg-Rademacker,Kevin B. Smith
• DOI: 10.1177/106591290305600210
• Date: 01 June 2003
• Published date: 01 June 2003
• Subject Matter: Articles

Academic educational policy analysis has been widely

used to justify prescriptive reforms of the public

school system and as a laboratory to seek confirma-

tion for broad theoretical frameworks such as public choice

and representative bureaucracy (e.g., Chubb and Moe 1988,

1990; Henig 1994; Smith and Meier 1995; Meier, Wrinkle,

and Polinard 1999; Maranto, Milliman, and Stevens 2000;

Witte 2000; Weiher 2000; Schneider, Teske, and Marschall

2000). A significant portion of this research is methodolog-

ically moored to education production function analysis, a

modeling approach that seeks to isolate the determinants of

school outputs using correlation or regression techniques

(Fortune 1993). Policy analyses using this approach, how-

ever, are notorious for producing inconsistent and contra-

dictory results on what does (or does not) determine school

performance (for surveys see Burtless 1996; Verstegen and

King 1998). As a number of theoretical issues important to

political science are now staked on such analyses, the disci-

pline has an important interest in this question: What

explains the variable and incongruent patterns reported in

empirical studies of school performance?

In this article we argue that a good deal of the inconsis-

tency is attributable to the conceptual problems created by

exporting the analytic framework of production functions

from economics and applying it to public policy and public

agencies. We argue these problems can be minimized

through a set of “second best solutions” that can be broadly

applied to policy studies with little need for any method-

ological re-tooling by researchers.

THE LOGIC AND PRACTICE PRODUCTION FUNCTIONS ANALYSIS

The production function framework that logically struc-

tures a good deal of empirical policy analysis in political sci-

ence originated in economics, and was designed for pro-

ductivity and efficiency analysis of private firms. The best

known formal treatment of the model is the Cobb-Douglas

production function:

X = f(L,K) = LaKb

Here, a firm’s output (X), is modeled as a function of L

(labor) and K (capital). The relationship between these

inputs and outputs is recovered through aand b, the param-

eters to be estimated. Given valid measures of output, labor

and capital, the model allows easy calculation of the mar-

ginal impacts of the factors of production. The output meas-

ure is assumed to be directly related to the overarching goal

of the firm. This goal is assumed to be profit maximization,

and profit (P) is defined as: P = cX - cL - cK, wherecrepre-

sents costs. Given this, the production function model

reduces questions of how to maximize a firm’s central goal

to a few straightforward calculations (specifically, the maxi-

mizing conditions for P are when P/L = 0, and P/K = 0).

This provides a parsimonious, theoretically derived analytic

framework to assess outputs and goal maximization for

firms (for an accessible primer see Zellner et al. 1966).

The basic logic, if not the mathematical elegance, has

long been exported to the rest of the social sciences. Typi-

cally economic theory is replaced with an alter nate frame-

work that intuitively “fits” the input-output logic of the

Cobb-Douglas model. In political science this usually means

substituting some unit of government for the firm, adopting

some measure of agency or policy output such as expendi-

tures or regulatory actions, and replacing labor and capital

with a set of inputs causally linked to outputs by some

throughput process undertaken at the unit of analysis. It is

also common to impose linearity restrictions on the input-

output relationship, thus making analysis more tractable

using basic multi-variate regression techniques.1

223

Money Only Matters if You Want It To?

Exposing the Normative Implications of

Empirical Research

KEVIN B. SMITH AND J. SCOTT GRANBERG-RADEMACKER, UNIVERSITY OF NEBRASKA AT LINCOLN

Political scientists have widely employed production function models as tools for theory confirmation and

policy prescription. Although an important part of a growing literature within the discipline, production func-

tion research frequently produces contradictory results, an inconsistency that raises questions about the role

of normative preferences in quantitative analysis. In this article we seek to explain the variation in the empir-

ical results of production functions research in education. While recognizing normative values may have some

influence in research, we argue this can be minimized if the conceptual and methodological weaknesses inher-

ent in applying production function logic to the public sector are recognized and addressed.

Political Research Quarterly,Vol. 56, No. 2 (June 2003): pp. 223-232

1A good example in political science is the “mainstream model” of state-

level policy outputs derived from Easton’s (1965) systems theory of pol-

itics. Here policy outputs are modeled as function of inputs such as

public opinion, state wealth, and partisan control of government. This