28 10 th IMEKO TC7 International Symposium June 30−July 2, 2004, Saint-Petersburg, Russia MEASURING INSTRUMENTS IN ECONOMICS Marcel Boumans Department of Economics, University of Amsterdam, Amsterdam, The Netherlands Abstract − The measurement set-ups for measuring economic phenomena are economic systems that cannot be controlled or shielded. Economists have to infer the desired quantitative facts about the phenomenon under study from the data generated by these systems. Accurate measurement results are therefore not obtained by adjustments and im- provements of the measurement system, but must result from building accurate models of these systems. By taking into account the background noise when modelling the phenomenon, the impossibility of control for bias and error is compensated. As a result, accuracy is obtained by first finding accurate representations of the relevant economic system incorporating both phenomenon and its environment. Secondly, these representations are applied in the methods for achieving precision of the measurement results. Because the unobserved facts about the relevant economic phenom- ena are inferred from data generated by the underlying system, measurement in economics is always associative in which the inferences are based on the models that function as representations of this association. Section two accounts for this kind of measurement. The principles of a broad range of modelling strategies in economics will be discussed in section 3. The general principles of making measurement results precise will be explained in section 4. Keywords: model, passive observation, ceteris neglectis 1. INTRODUCTION In empirical economics, models are built to pro- vide facts about phenomena. Though phenomena, like business cycles, GDP and unemployment, are the objects of explanation and prediction of economic theories, these theories usually do not generate quanti- tative facts about them. For example, theories tell us that capitalist economies give rise to business cycles, but not the duration of recovery. These facts about economic phenomena are not directly observable, but nevertheless have to be converted from data, i.e. ob- servations (see [1] for a elaborate discussion of the distinction between data and facts about phenomena). If data are arranged in a considered way, meeting specific requirements, they provide reliable quantita- tive information about phenomena. In economics, models function as such devices. In other words, the measuring instruments of economists are models, that are located on the theory-world axis mediating be- tween facts about phenomena and data, see Fig. 1. The dotted line in Fig. 1 represents the indication that theories do not provide (quantitative) facts about phe- nomena. Theory ↓ Phenomenon Facts about the phenomenon ↑ Measuring Instrument ↑ Data Fig 1. Position of measuring instrument on theory-world axis There is, however, one crucial feature of models that has profound consequences for achieving accurate measurement results. Namely, models are ideal sys- tems, and so all kinds of material interventions to make measuring instruments reliable, such as control, shielding and insulation, are not possible for mathe- matical objects. To function as an accurate measuring instrument, a model should include a representation of an as far as possible invariant relationship between facts and data. Though invariance is hard to find in an economic world of constant flux, this paper will ex- plore some strategies of finding representations of invariant relationships. These strategies are based on works of economists and econometricians who sug- gested influential solutions to this problem of invari- ance. Before these works will be discussed, we first, in section 2, have to clarify the concept of a model as an ideal measuring instrument. Therefore, we will use Heidelberger’s [2] correlative interpretation of the representational theory, which is based on Fechner’s correlational theory of measurement. In Heidelber- ger’s work a correlation is linked to a measuring in- strument ensuring the invariance of this relationship. Generally for material instruments this can be done by insulation, or in other words by taking care that ceteris paribus requirements are fulfilled. In economics, we often cannot create ceteris paribus environments and so have to search for invariant relationships in the open air. Do constellations in nature exist that can be used as measuring instruments and how do they look like? The works of Trygve Haavelmo [3], Herbert Simon [4], Milton Friedman [5], and Robert Lucas [6]