Gerhard Tintner, The Use of Mathematical Methods in Econometrics and Economic Statistics (1954)

Gerhard Tintner’s journal article surveys the mathematical resources required to turn economic theory into empirically estimable relationships and policy-relevant knowledge. Its structure moves from a classification of economic models through data collection and specifically econometric difficulties to probability, estimation, confidence limits, and hypothesis testing. Mathematics is presented as a means of connecting theoretical assumptions, observations, and practical decisions—not as an independent guarantee of economic knowledge.

Econometrics[^641-1] may be defined as an endeavour to use the methods of mathematical economic theory and of modern mathematical statistics in order to accomplish two goals: to find numerical values for the postulated economic relationships and to verify economic laws and regularities.

These two purposes organize the survey. Tintner distinguishes economic statistics, which collects and presents data, from econometrics, which estimates and tests relationships, while arguing that their practitioners need closer intellectual cooperation. Data collection must anticipate analytical uses; statistical analysis must understand the economic structures it seeks to measure.

The first major section classifies models along two axes: partial versus general equilibrium, and static versus dynamic analysis. The Swedish demand for butter illustrates static partial equilibrium. Estimated elasticities express how demand changes with butter prices, margarine prices, and income, holding other factors constant. Their usefulness to producers and governments depends on precisely this conditional interpretation. Supply, cost, production, and utility functions extend the same approach. Algebra and calculus provide its basic techniques, while game theory introduces strategic interaction and linear programming identifies productive combinations under specified objectives.

General equilibrium shifts attention from individual markets to interdependent economic systems. Haavelmo’s consumption model connects the marginal propensity to consume with an investment multiplier and possible government responses to deficient private investment. Leontief’s input-output analysis instead represents flows among economic compartments, using constant production coefficients to trace system-wide consequences. Tintner thus distinguishes aggregate income relationships from sectoral interdependence rather than treating every economy-wide model as interchangeable. Dynamic models add expectations, past experience, and time-lags. Whitman’s steel-demand equations and Tinbergen’s seventy-variable business-cycle model demonstrate why such analysis requires differential, difference, and integral equations, and sometimes more advanced methods.

The section on economic statistics brings these mathematical ambitions back to the production of evidence. Tintner regrets reliance on administrative by-products, such as income-tax records, and advocates better coordination between collectors and users of data. Sampling offers frequent information at lower cost than complete censuses, but the appropriate design depends on what the econometrician ultimately needs to establish. Reliable inference therefore begins before estimation, with the organization of observations.

Aggregation and identification then emerge as problems that are neither purely economic nor purely statistical. Aggregation makes vast theoretical systems empirically manageable, but raises the index-number question of how combined prices and quantities can preserve relevant properties of the underlying relationships. Identification asks whether available observations can disclose the structural equations and coefficients specified by a model.

If the equation is not identified, then no statistical analysis of the data can reveal the desired structural relationships or give us estimates of the structural parameters, which we desire to know.

This is the article’s strongest methodological limit on statistical technique. In Tintner’s demand-and-supply example, observed price and quantity alone do not distinguish the two relationships; introducing income into demand and costs into supply supplies the additional structure needed for identification. His distinctions among endogenous, exogenous, and predetermined variables make explicit how estimation depends on economic assumptions, including assumptions about what lies outside the model.

Time series introduce another obstacle: successive economic observations cannot ordinarily be treated as independent random events.

It is not possible to utilize without change the statistical methods which have had great success in other fields.

Trend removal, differencing, autoregressive transformations, and stochastic-process analysis offer responses, but short series may fit several competing models equally well. Transformations themselves must be selected from limited evidence. Tintner consequently distinguishes the applicability of general principles of inference from the uncritical transfer of particular statistical procedures.

The final sections explain competing conceptions of probability and the practical roles of estimation, confidence limits, and tests. Frequency-based probability concerns repeated outcomes; probability as rational confirmation concerns the evidential standing of propositions, although Tintner considers the latter insufficiently developed for most econometric applications. Examples involving labour productivity and meat demand illustrate numerical estimation through maximum likelihood and least squares. Confidence limits and Neyman–Pearson tests qualify numerical findings through repeated-sampling uncertainty and the distinction between rejecting true hypotheses and failing to reject false ones.

In a subject matter with so many conflicting ideas the best attitude is perhaps an eclectic one.

This methodological pluralism remains selective. Tintner closes by questioning Wald’s decision theory: economic applications may lack defensible numerical loss functions, while minimax reasoning implies a highly conservative policy. The article’s relevance lies in its integrated account of mathematical possibilities and empirical constraints. Useful econometrics requires appropriate models, purposeful data collection, identifiable relationships, and uncertainty-conscious inference; sophisticated technique cannot substitute for any of them.