LSE alum Sebastian Petric argues that the prevalence of environmental, social, and governance (ESG) in shaping markets, businesses and institutions should prompt us to redesign the models used to predict financial crises in emerging markets.
Institutions are the “rules of the game in a society,” as Douglass North put it, and institutional factors are among the key predictors for crises in emerging markets. Over recent years, environmental, social and [corporate] governance (ESG), a set of standards used by socially conscious companies to monitor policies, practices and investments in these three areas, has taken over as the more prominent focus in finance and in the literature on early warning signals of financial crises. ESG shares similarities to institutional variables in that ESG or “socially responsible” principles have become a dominant trend which businesses and markets must operate within. The first question I would like to discuss in this blog is whether ESG factors are indeed the same as institutional variables. Governance is among the key institutional variables (please see for instance the Worldwide Governance Indicators for more details), social variables are also very similar to institutions, but what is very much an innovation to the institutional literature are environmental factors.
Generals are always prepared to fight the last war, i.e., conclusions are drawn from and apply to the most recent conditions, and these are used to anticipate the next crisis. Could it be different this time? I believe yes, environmental factors could be among the key explanatory factors for crises in emerging markets, and thus, should be included in early warning systems. The most common model used in these systems are logistic regressions which is a statistical model used to calculate probability based on a number of variables. The problem with this econometric technique is that explanatory variables are linked to historic crisis episodes (e.g., the 1997 Asian Financial Crisis) and the crisis exposure is estimated with the help of a technique called maximum likelihood. However, environmental factors were not among the key explanatory variables in these crises and thus, this econometric method will struggle in predicting future crises. That being said, every early warning system analysis should still start with this econometric method as although imperfect, it provides insights into the drivers of past crises which could be relevant to future crises.
Hence, the second point I would like to discuss is the methodologies which can be applied to predict financial crises in emerging markets going forward. As I argued above, environmental factors are an addition to the classical institutional variables which are used to predict crises. What innovations can be introduced to improve the accuracy of crises predictions? Machine learning algorithms is one convincing answer to the above, in particular, unsupervised learning techniques. In these techniques, there are no response variables (i.e., dependent variables), which spares us of establishing the historic link between crises and explanatory variables in the first place. One such technique, which can be particularly useful is clustering. In clustering, variables are grouped according to certain features and hence, these could be clustered quantitatively according to environmental risks.
To conclude, institutional variables have historically been among the key predictors for emerging market crises. However, environmental risk factors are becoming an increasingly urgent factor in early warning systems. Thus, we must find innovative ways of incorporating environmental risks into our methods of calculating early warnings of crises in emerging markets.
The views expressed in this post are those of the author and in no way reflect those of the International Development LSE blog or the London School of Economics and Political Science.
Photo credit: SAISPerspectives.com.
