Thursday, April 08, 2021

No Finshots, it is not the data

Today, RBI kept the interest rates the same after a 3-day deliberation. 

Finshots is a email service/app that writes short journalistic pieces on the.. well, financial sector. 

In today's story, Finshots hypothesises that : 


And an oft-cited criticism of the Reserve Bank of India is that it consistently fails to forecast inflation and growth rates accurately. According to the critics, this is a rather serious transgression on the part of the bank, mainly due to the fact that these numbers are then used to justify the monetary policy calibrations [changing interest rates etc]. How could the collective wisdom of the Monetary Policy Committee, the governor, the deputy governors, the modelling experts and other technical advisors all be so woefully inadequate to deal with what seems like a rather straightforward problem?

To answer that question we must first look at the data that’s used to predict these future outcomes. As Duvvuri Subbarao, the former Governor of the RBI points out — “The Reserve Bank operates within the universe of knowledge available in real-time, and that universe is largely shaped by data. If the data are reliable and available in good time, policy response can be accurate and confident. But the Reserve Bank is oftentimes wrong-footed because of the questionable quality of data.”


Let me tell you a story. 

Naresh wanted to predict the rainfall this year. So, he took data from the last 20 years' rainfall - all 20 chaumasas with their comings and goings and everything in between. 

Then, he used a predictive index to say which day of the year will see how much rain. 

The statistical model predicted a deluge. People took precautions, businesses delayed new launches until after the deluge, and so on. 

But, that quote - Man proposes, God disposes, was not created in vain. The rains were normal, or below normal. People suffered that and criticised Naresh. 

A good friend came to meet him and asked him how he had configured his predictive model. Naresh said - the highest rainfall received on that date in any year was used to create the prediction of the expected rainfall in this year.  

Obviously, Naresh had all the data. He simply did not put the right analysis to that data. 

But why would anyone make such a huge mistake? Why not use something as simple and obvious as mode, or even mean rainfall? 

Because, Naresh had taken that model from a Cherrapunji based scientist. There, they prepare for excessive rains. So, they predict based on the highest they have ever got. Because the variance is not too high, it doesn't matter. They prepare for the worst and it works best for them. 

When we see that the model was created by a scientist in Cherrapunji, what appears to be absolute lack of reason, makes complete sense. 

End of story. 


Why is that story relevant here? 

Because, the financial models we use to predict our numbers are created by economists who work in western economies. For over 200 years, everyone has been trying to fit our economy and its variables to a single universal format, guided by research in macro and behavioural economics. This research has been conducted in largely homogenous economies, with a defined set of cultural and social parameters. 

My hypothesis is, that if we want to understand the economy of India and how it interacts with macro economic policy variables, we have to understand how THIS economy functions. The government has often criticised the Indian trading community as being less than transparent. But is the government also not guilty of remaining ignorant about how the trading classes actually function, and then to create rules and laws that keep that in mind? 

The issue is not just with the data. The issue is that we have not invested in research to understand the model within which we operate. We report on numbers whose formulae are imported and have limited application in a complex economy like India. 


1. When the high denomination currency notes were banned overnight, no one expected the poorest of the poor to be so inconvenienced. BUT, they were. The preliminary, back of envelope calculation I did back then indicated the massive cash that flows through the bottom of pyramid. But because all this cash is completely undocumented, we cannot even include that data in our models, much less know what to do with it. 

2. Borrowing to spend on weddings and funerals is not listed anywhere as rational economic behaviour. No model of economics accounts for this. Yet, for at least 1000 years, Indians have borrowed routinely for weddings and funerals (if they are poor). What makes this even more perplexing is that this is the same society that gives 10% to charity and has among the highest family savings rates in the world. Why, then, do they borrow for weddings and funerals? (And now, for education) How does this indebtedness impact their disposable income and consumption? Their total income? Their nutrition? For how long does the impact last? 

3. How much does a country like India really need to spend on welfare medical facilities for all its people? What are the real medical needs of people who live every day with snake bites and dysentery?

4. In 2020, as spending was just not picking up, the government slashed savings rates. they hoped that this would lower credit rates, leading businesses to pick up cheap credit, and that would lead to economic activity. But that did not happen. Why? Economic activity picked up around Diwali. Why? I wrote about this in another post - the Indian does not borrow when there is an income squeeze. If the government had done the exact opposite - increased savings rates, that would have led to a perception of income in the mind of the consumer and would have spurred retail economic activity. Diwali is a time when retail activity has to go up. 

5. RRBs were created, at much cost, to replace the mahajans. Co-operatives were created to ensure that small Indian land holdings do not lead to low productivity. Neither worked. Vinobha Bhave, on the other hand, went from village to village and in a land-crazy country like India, made something like Bhoodan happen. But 7 decades later, Cooperatives still dont have large land packets. And 45 years later, the Mahajan still rules rural credit. The thing to do, was not to create RRBs. To get rid of the Mahajan system, one had to understand WHY the Mahajan system works and then decide whether the system needs fixing or replacement. 

The same thing goes for every economic system in India - the labour market, the sheer vastness of the unorganised sector, in all 3 sectors of the economy, the impact of the social structure on economic activity. In everything, we need to understand the model to weave it into our algorithm. 

The fault is not with RBI alone. Most economic reporting of India leaves out India (Sure, you can call it Bharat and snigger, but well, where's the money, honey?) 

It is written by people educated in Keynesian economics, read by people being educated in Keynesian economics, and criticised by people who base their criticism on fancy jargon and models of... you guessed it. 

India's policy making, macro economic indicators reporting, and indeed, our education of economics, cannot be conducted without understanding the native economic models of India.

No comments: