Introduction:

The overall goal of this paper is to scrutinize the way of dealing with the matter of causation in social sciences. We realized that it is rather common to infer causation from empirical data in social sciences; and we wonder how causal inferences are commonly made in social sciences and whether or not they are valid. To approach these general questions we will examine an important contribution to growth economics: "Why do some Countries produce so much more Output per Worker than others?" by Robert E. Hall and Charles I. Jones (1999). Applying the method of instrumental variables, Hall and Jones argue in favor of revealing institutions as a fundamental cause of long-run economic growth. The impact of institutions on a country’s wealth has not been placed in question in more recent growth economics; but Hall and Jones have been criticized for their approach of inferring just this causation by researchers such as Acemoglu et al. (2005) and Eicher and Leukert (2009). Our contribution will be to call attention to the need of dealing with the matter of causation in a more cautious and thoughtful way. We will do so by exposing major weaknesses both in the way Hall and Jones come to their conclusion and in the causal model they assume. To follow the questions described above, the first chapter is devoted to the paper by Hall and Jones (1999). We will trace how the authors arrive at the conclusion that institutions do indeed have causal relevance for economic growth. To this end, we will describe their hypothesis, the statistical method they apply to come to reach this result, and the causal structure they state. In the second chapter we will exercise criticism. Hall's and Jones's analysis is based on certain assumptions concerning the associations between exogenous and endogenous variables, as well as on latent and observable variables. These assumptions are necessary for the statement they seek to make. We now want to argue that other associations that are no less reasonable, but whose hypotheses are less convenient, should be considered as well, if Hall’s and Jones’s intention of analyzing for a causal relationship is to be deemed honest. We will then comment on their data base and show some interesting findings concerning the significance of the variables they use for their study. We will then revisit the critical points we have made in the previous chapters and include them in an adapted graphical model. In the fourth part of the paper we will test the causal structure Hall and Jones argue for. We will do this by using Tetrad, a program for creating and testing causal and statistical models. Working with the same data as Hall and Jones and applying two different algorithms, the PC-Algorithm and the FCI-Algorithm, the program will produce interesting results that support our critical deliberations while falsifying not only Hall’s and Jones’s structural model, but even their major hypothesis. In the last part of the paper we will emphasize our findings. We will reach the conclusion that Hall and Jones analyze and state an economic association that is definitely interesting and commendable, but that at the same time they do not pay sufficient attention to the difficulty involved in finding causal structures. They deal hastily and imprudently with the matter of causation, and thus make a strong claim but using weak instruments.