Wrong path

Designing a growth strategy is like getting to the peak of a mountain that is covered by clouds, states economist Francisco Rodríguez in a paper for the International Poverty Centre in Brasilia. You have a rough idea of the topography but the details are hard to make out. Linear growth regression analyses, which are the standard in applied research, suggest that all mountains look alike – smooth and regular, like a pyramid. And that is why they are so tempting for politicians: they give the impression that there is a single recipe for economic growth, applying to all countries. Rodríguez adds that this assumption also makes economic models a dangerous tool, as the real economy is much more complex than regression analyses imply. Oversimplified growth models sometimes lead politicians on to the wrong path, leading to damage rather than an economic high ground.

Regressions set out to establish cause-and-effect relationships between variables, for instance, between the openness of a country’s economy and its growth rate. Such analyses have three weaknesses, Rodríguez writes. First of all, the imputed causality is questionable. Regressions show that trade openness and growth impact on each other. But the direction of causality is not clear: while trade may accelerate growth, it is equally possible that growth may encourage policymakers to lower trade barriers. What the analysis identifies as cause and what it recognises as effect depends on the assumptions made by the economist.

Secondly, the variables selected in regressions often produce results that are of little use to politicians. What good is it to a government to know how a particular degree of openness – the ratio of exports plus imports to GDP – impacts on growth? Politicians want to know what will happen to growth if certain tariffs or other trade barriers are lowered. According to Rodríguez, however, there is no measurable correlation between tariffs and growth.

Thirdly, Rodríguez points out that results are often not very reliable. The inclusion or omission of certain variables and changes in the size of the database can impact significantly on the results of a regression.

Rodríguez concedes that growth theorists can mitigate these weaknesses by making judicious adjustments. But one problem still remains: “The workhorse growth regression embodies a particular vision of the world that assumes, implicitly, that the same model of growth is true for all countries.” In that sense, the assumption that the openness of a country is a cause of growth will not vary, regardless of how open, how economically diversified and how poor (or rich) it already is.

Rodríguez believes more dependable results are furnished by so-called nonparametric analyses. Unlike conventional regressions, these are mathematical analyses that are based on variables which are not specified a priori and do not depend on fixed assumptions on the correlations between them. The structure of nonparametric models is developed from the data, rather than defined in advance. Thus, in Rodríguez’ view, these models serve to capture real-world complexity more accurately than standard regressions.

Another way to avoid the pitfalls of standard regressions, Rodríguez says, is to apply the growth diagnostics approach developed by economists at Harvard University. This method starts with a given problem – e.g. low investment – and progresses through a checklist of possible causes and the parameters that give rise to them. This, Rodríguez claims, enables faulty stimuli in an economy to be identified and corrected.

Anyone who wants to climb a cloud-covered mountain needs to pay close attention to the ground underfoot. And anyone wanting to get an economy into gear needs to study its structure. According to Rodríguez, whatever simplistic growth models suggest, it is illusory to think that one size can fit all. (ell)