James A. Chalfant

Accounting for Changes in Tastes

Health concerns are thought by many to have shifted consumption away from red meats, though econometric evidence is mixed. Testing for structural change is difficult, especially when one time series is used for both estimating demand equations and testing their stability. Specification errors may suggest a shift where none has occurred. Using nonparametric demand analysis, we find that meat consumption patterns in the United States and Australia can be explained using only relative prices and expenditures. Only imposing particular functional forms can reverse the conclusion, suggesting that specification errors in econometric demand studies can account for findings of taste changes.

A Globally Flexible, Almost Ideal Demand System

The almost ideal demand system is combined with the Fourier expenditure system. Subject to the assumption that preferences are of the price-independent, generalized-logarithmic class, the resulting demand system has the desirable features of each of its components. Aggregate demand equations are consistent with preferences of a representative consumer, and consistent estimates of elasticities are obtained for all observed prices. Application of the new demand system to U.S. consumption of meats and fish reveals that it fits the data well and that the restriction to the usual specification is rejected.

Marketed Surplus under Risk: Do Peasants Agree with Sandmo?

Using a newly defined notion of aversion to income risk, the behavior of the marketed-surplus producer under price risk is characterized. Unlike the familiar case first examined by Sandmo, output depends on both ordinal preferences for goods and on risk attitudes. Conditions are found that yield an output level under risk that is smaller than under certainty. If these conditions do not hold, both risk and risk aversion may have a positive effect on output. Implications for econometric studies of risk attitudes are considered and illustrated with an example. Finally, we examine the effect of uncertainty on the peasants long-run equilibrium.

The Silence of the Lambdas: A Test of the Almost Ideal and Rotterdam Models

During the past decade, the Linear Approximate (LA) Almost Ideal Demand System and the Rotterdam model have been by agricultural economists as the demand systems of choice in most applications. The apparent explanation is that the two models are both (second-order) locally flexible and compatible with demand theory, they have identical data requirements and are equally parsimonious with respect to parameters, and both are linear in the parameters. While the two models are thus equally attractive in most respects, and indeed appear very similar in structure, they lead to different results in some applications. This article develops a test of each against the other. In an illustrative application to U.S. meat demand, the Almost Ideal model is rejected while the Rotterdam model is not.

Demand Response to Advertising in the Australian Meat Industry

The implications of model specification choices for the measurement of demand response to advertising are examined using Australian data. Single-equation models versus complete systems and alternative corrections for autocorrelation are evaluated. Competing advertising efforts by two producer bodies are included. Across all specifications, the evidence on advertising effects is fairly consistent. In the preferred model, the only statistically significant effects of advertising are for Australian Meat and Livestock Corporation advertising (of beef and lamb) on the demand for beef (positive) and on the demand for chicken (negative). Australian Pork Corporation advertising does not have any statistically significant effects. Copyright 1996, Oxford University Press.

Evaluating Prior Beliefs in a Demand System: The Case of Meat Demand in Canada

An almost ideal demand system for meats is estimated using Canadian data. A Bayesian approach is used to impose inequality restrictions on substitution elasticities, via Monte Carlo integration and importance sampling, in order to conform with prior beliefs about curvature and monotonicity restrictions and substitution relationships. Results are more consistent with the concavity and monotonicity restrictions from demand theory than with the added restriction that all meats are substitutes.

Estimating substitution elasticities with the Fourier cost function : Some Monte Carlo results

Abstract The Fourier flexible form possesses desirable asymptotic properties that are not shared by other flexible forms such as the translog, generalized Leontief, and generalized Box-Cox. One of them is that an elasticity of substitution can be estimated with negligible bias in sufficiently large samples regardless of what the true form actually is, save that it be smooth enough. This article reports the results of an experiment designed to determine whether or not this property obtains in samples of the sizes customarily encountered in practice. A three-input, homothetic version of the generalized Box-Cox cost function was used to generate technologies that were oriented in a two-dimensional design space according to a central composite rotatable design; the two factors of the design were the Box-Cox parameter and a measure of the dispersion of the substitution matrix. The Fourier cost function was used to estimate the substitution elasticities at each design point, and the bias at each point was estimated using the Monte Carlo method. A response surface over the entire design space was fitted to these estimates. An examination of the surface reveals that the bias is small over the entire design space. Roughly speaking, the estimates of elasticities of substitution are unbiased to three significant digits using the Fourier flexible form no matter what the true technology. Our conclusion is that the small bias property of the Fourier form does obtain in samples of reasonable size; this claim must be tampered by the usual caveats associated with inductive inference.

The Changing Empirical Definition of Money: Some Estimates from a Model of the Demand for Money Substitutes

Interest-bearing checkable deposits are examined to test whether they should be included in measures of the U.S. money stock. Both Divisia and traditional simple-sum aggregates are constructed on the basis of tests for weak separability in a model of the demand for financial assets. Using nonparametric demand analysis, we find that several groups of assets are compatible with aggregation theory. We find empirical support for a narrow measure consisting of the components of current MIA. In tests based on a St. Louis equation and in terms of controllability, a Divisia aggregate performs better than the simple-sum MIA measure.

Incorporating demand shifters in the Almost Ideal demand system

Abstract Intercepts of share equations generally include demand shift variables. In the Almost Ideal demand system and related models, this results in estimates that depend on units of measurement. Solutions to this problem are identified and discussed.

The Incidence of the Costs and Benefits of Generic Advertising

Generic advertising is important and controversial. In 1990, commodity organizations in the United States spent over