What is Philosophy of Econometrics?
The philosophy of econometrics is still a young subdiscipline that needs a more detailed clarification. Econometrics claims to be the empirical basis of economics, since it can explain economic phenomena with mathematical, predominantly statistical methods. To do this, economics must pass through as inductive science. Because only inductive conclusions are knowledge-expanding. If a discipline is to count as empirically founded science, measurable data is required. Philosophy, on the other hand, is the study of general, fundamental questions about existence, knowledge, values, reason, consciousness, and language. So the core of philosophical work consists of asking general questions. An essential branch of philosophy, the philosophy of science, deals with fundamental questions in other individual sciences. This includes, among other things, the question of what adequate, scientific explanations are. Thus, the question of whether there is causality at all exists since Aristotle as a question of philosophy. So what questions can be relevant to philosophy of econometrics if one discipline is provided with concrete data, while the other discipline examines problems of a general nature? This question is accompanied by further considerations. Can we explain with data? What can we explain if this is true? There is a common belief in the philosophy of science that we must find causal mechanisms to adequately explain phenomena. How can data help us? Measurable data represents isolated observations and say nothing about causal mechanisms. The basic idea for the main tool of econometrics, regression analysis, is to relate measurable data. But in which relationship are they set? Regression analysis – I’m just using the simple version here – points to the functional dependency of variables:
However, a functional dependence is not a causal dependence because we need the exact direction from cause to effect for them. This is precisely the problem to which the philosophy of econometrics devotes itself: causality is not provable empirically. It must therefore be formulated in a theoretical preliminary work. On the other hand, measurable data, which is positively related in a regression analysis, only reflects a correlation. Not infrequently, the assumption is made that we can at least expect a causality in positive correlations. This is accompanied by some problems. For a regression analysis, assumptions must be made that can ignore important data. Justifiably, therefore, the question must be asked whether we can even explain with regressions. Nevertheless, they are used for practical purposes, as some examples will show during this talk. The intention is to draw attention to two problems. On the one hand, it should be pointed out that you consider many events to be „explained“, even though the concept of explanation is much more complicated than you thought. On the other hand, it aims to show how the unobserved difference of causality and correlation can open the space for false assertions and manipulation. What are the consequences if we consider phenomena to be explained, even though the principles of these explanations are not suitable for doing so?