Gerhard Tintner, Econometrics (1953)

Gerhard Tintner’s journal article presents econometrics as a method for turning theoretical economic relationships into conditional numerical knowledge. It moves from the construction of mathematical models through structural estimation, identification, and statistical inference to the tentative evaluation of policy. Its central argument is that economic theory supplies indispensable relationships but cannot itself determine their numerical magnitudes; statistical methods can estimate those magnitudes only under assumptions whose validity must remain open to scrutiny.

Econometrics consists in the application of a specific method to economic problems:² the use of mathematical economic theory and of mathematical statistics in the field.

The conjunction matters: measurement requires theory, while theory needs empirical specification. Tintner begins with the maximizing consumer and firm, whose behavior can be represented by systems of equations. Yet maximization does not determine whether a demand relationship should be linear, logarithmic, or otherwise specified. Different functional forms imply different empirical properties, including whether elasticities remain constant.

Economic theory can only give us some very general properties of these functions, but not their specific form⁶.

This gap between theoretical restrictions and numerical specification organizes the article. Tintner distinguishes static from dynamic models and microeconomic behavior from macroeconomic aggregates. Static models often have to be applied to genuinely dynamic observations because adequate theories of expectations and anticipations are lacking. Aggregation presents another unresolved difficulty: national income, consumption, and employment make empirical analysis manageable, but their relationship to individual decisions cannot simply be assumed. These are limitations of the theoretical framework, not merely problems that improved statistical technique will remove.

Structural coefficients provide the bridge between models and economically interpretable results. Price and income elasticities, marginal productivity, and the marginal propensity to consume express what a model implies about changes in economic behavior. Tintner illustrates them with American estimates for beef demand, labor productivity, and consumption. His repeated qualifications are essential to their meaning: an estimated response applies only if the model and estimation assumptions hold, circumstances resemble the observed period, and other relevant conditions remain unchanged. The figures are conditional descriptions of relationships, not unrestricted economic constants.

Estimation also requires a stochastic model. Tintner distinguishes errors in variables, comparable to observational errors, from errors in equations, arising when relevant influences are omitted. A demand equation cannot realistically contain every price that a complete Walrasian system would require. Its incompleteness therefore enters the statistical specification. His meat-market example produces similar demand elasticities under the two error assumptions, but this agreement does not erase the conceptual distinction. Assumptions about distributions and dependence must still complete the model.

Identification precedes the choice of an estimator: the investigator must establish whether the desired structural relationships can be distinguished at all.

By investigating the identification of the equations in our model we want to ascertain, if the estimation of these structural parameters would be possible at all, quite apart from all the complications introduced by sampling.

In the agricultural demand-and-supply example, equilibrium observations alone do not separate the two relationships. Identification becomes possible when income enters demand and costs enter supply, with both treated as exogenous. Tintner stresses that genuinely exogenous economic variables are rare; treating a variable as external is often justified only relative to a limited market. Predetermined variables offer another route. His dynamic corn-market model uses previous prices and stocks to identify relationships involving current demand, supply, and price, while introducing the additional statistical difficulties of stochastic difference equations.

Once equations are identified, estimation can proceed, frequently through maximum likelihood. Tintner then moves beyond point estimates to confidence or fiducial limits and hypothesis tests. An estimated meat-demand elasticity accompanied by an interval indicates uncertainty that a single coefficient conceals. But those limits, often approximate, remain dependent on the model’s economic and stochastic assumptions. Statistical precision is therefore not an independent guarantee of theoretical adequacy.

Hypothesis testing makes econometrics relevant to disputes within economic theory. Tintner discusses tests of whether labor productivity is zero and whether behavioral functions are homogeneous of degree zero in monetary magnitudes. His English labor-market example does not reject dependence on real rather than money wages, a result he places against Keynesian assumptions. Nevertheless, non-rejection remains conditional, rather than establishing the model conclusively. Tests supply tentative grounds for comparing theories, while the general problem of choosing between competing models remains unresolved.

The closing discussion broadens this caution to probability and inference themselves. Tintner considers Carnap’s ideas promising but not yet applicable to complicated econometric problems, notes divergences between Fisher and Neyman–Pearson inference, and questions the suitability of Wald’s decision-theoretic approach for social science. Econometrics nevertheless retains a practical role:

Econometrics cannot of course, formulate the goals of economic policy.

Ethical and political judgments determine the ends; estimated structural relationships may help evaluate the means. Tintner’s concluding meat-market example projects effects of taxation and subsidy on consumption and prices, then reiterates that its assumptions may not be fully realized. The article’s significance lies in this disciplined balance: econometrics extends economic reasoning into numerical estimation and empirical testing, but its policy guidance remains provisional because theoretical adequacy, identification, stochastic specification, and changing historical conditions govern what the numbers can legitimately mean.