A puzzling data revision

Ordinarily, official statistics get revised because at first, provisional estimates are released, and when the full data filings come in, then improved estimates are put out.

In the case of RBI's data about RBI's trading on the currency market, such data revisions should ordinarily not arise.

But yesterday, data released by RBI modified the previous information that had been put out about RBI's trading on the currency market. Earlier, trading in June had been claimed to be 0. Now it shows purchase of $370 million and sale of $260 million. Earlier, trading in September had been claimed to be 0. Now it shows a purchase of $260 million. I wonder why this data revision took place.

The newest data - for October - shows a purchase of $450 million on the spot and $450 million on the forwards. At a time when rupee trading is estimated at above $40 billion a day (worldwide), it is hard to see how such a small scale of trading can generate a significant impact upon the price; so I wonder what is going on in terms of the rationale and the thought process.

Interesting readings

One of the big impediments to India's integration into the world economy is xenophobic visa rules. There is some progress in the pipeline: visa on arrival has been working from Jan 2010 onwards for visitors from Finland, Japan, Luxembourg, New Zealand and Singapore. A nice touch here is that India did not get stuck on issues of reciprocity; this is unilateral liberalisation.

Watch this talk by Steve Coll.

Mature treatments of the Niira Radia wiretaps : Sail Tripathi in the Mint, Pratap Bhanu Mehta in the Indian Express.

Anil Padmanabhan in Mint on the question of corruption, and Sevanti Ninan on the media response to the tapes.

In search of America's liberty and India's dharma by Gurcharan Das in the Times of India.

A rumination by Vikram Doctor, on the need to shift focus in Bombay from the West to the East.

Sam Geall on the problems of Chinese science. Some of these problems are found in India also.

With corruption scandals galore, what India needs most is competent and clean government. SEBI continues to soldier on: see the recent order on bond issues by Sahara. Or if you don't have appetite for the full text, here is a precis by V. Umakanth. Everyone interested in Indian finance should read a few orders of Bhave's SEBI every year: they give you fresh insights into how the interplay between law and regulation works.

Tamal Bandyopadhyay in Mint with his sense about the extent of corruption in Indian banking.

How do foreign capital flows behave around elections, on voxEU by Emmanuel Frot and Javier Santiso.

Currency warriors should consider India by Sebastian Mallaby in the Financial Times.

A. K. Bhattacharya writes in the Business Standard about fresh thinking on Indian Railways from an unexpected source.

Huang Yiping on voxEU has a story from China which is similar to what we often see in India: the use of microeconomic tools to go after macroeconomic problems.

In the footsteps of Gandhi, Mandela and Havel, by Ma Jian, on Project Syndicate. Unveiling hidden China by Christian Caryl in the New York Review of Books.

Good-bye to Dubai by Joshua Hammer in the New York Review of Books.

Robert Messenger looks back at Dien Bien Phu.

Richard Boudreaux in the Wall Street Journal about Russia's Parliament accepting Stalin's responsibility for the Katyn massacre.

Kenneth Rogoff on the Euro.

A tale from the frontiers of public administration. The Australian government has announced a competition to forecast the behaviour of traffic on Sydney's M4 freeway. This illustrates three themes. The first is that of better living through science: the attempt at using statistical analysis to shape public administration. The second is the unique value of public domain databases. The third is the importance of harnessing brainpower out there in innovative ways: through openness of data and through the competition.

Trailhead by E. O. Wilson. As I read it, I was astonished at the way in which knowledge gleaned from hundreds of research papers has been stitched into a compelling story.

A club of 19

What binds this club of 19 countries: China, Russia, Kazakhstan, Colombia, Tunisia, Saudi Arabia, Pakistan, Serbia, Iraq, Iran, Vietnam, Afghanistan, Venezuela, the Philippines, Egypt, Sudan, Ukraine, Cuba and Morocco? Answer. Am I glad India is not in this club!

Alternative stock market indexes

I saw this interesting article about the mind-share of Nifty as opposed to the BSE Sensex. It is by Samie Modak and Muthukumar K. in the Financial Express.

The NSE data for June 2010 shows that Nifty futures have peaked at Rs.0.36 trillion of notional turnover in a day (27 Jan 2010) and Nifty options have peaked at Rs.0.89 trillion of notional turnover in a day (24 June 2010). Nifty has shaped up as one of the big contracts by world standards. It is interesting to go back and read the original paper. Those were interesting times. Looking back, it seems obvious that Nifty would dominate the derivatives market, but at the time, the outcome was far from clear.

This made me look at data on risk and reward of the alternative indexes. I start from the first data for Nifty Junior, which takes me back to 21 February 1997, thus giving data for 13.7 years.

	Mean	Volatility	Ratio
Nifty	12.99	26.37	0.4926
BSE Sensex	12.68	26.92	0.4711
Nifty Jr.	18.16	32.38	0.5608
CMIE Cospi	17.40	27.23	0.6391

Nifty and the BSE Sensex are a lot like each other.

The real surprise is Nifty Junior: Merely moving down from rank 1-50 to ranks 51-100 has given an enormous juice in the return and in the reward-to-risk ratio. But the volatility of Nifty Junior is also higher.

The CMIE Cospi index has roughly 2800 stocks today, and represents the broad market. It includes the Nifty Junior stocks and a host of other smaller stocks. But unfortunately, these numbers are not comprabale with the other three in that it includes dividends while the other three do not. With this combination of high diversification (giving a low volatility), small-cap stocks (which helps returns) and inclusion of dividends (which helps returns), it is not surprising that it scores the best reward-to-risk ratio.

In my mind, most of the claims of out-performance by active managers in India are purely about being invested in the non-Nifty space. Nifty Junior ETFs are easily accessible and I get surprised that more people aren't putting this into their investment strategy.

A more efficient piece of code

CMIE's firm databases use a fine-grained product code to identify each product. Each firm is also allocated to a product code based on its predominant activities. I like to reconstruct a coarse classification out of this that suits my tastes. I do this using this R function:

cmie.14.industries <- function(s) {
  values.8 <- c("Food","Textiles",
                 "Chemicals","NonMetalMin",
                "Metals","Machinery",
                 "TransportEq","MiscManuf",
                "Diversified","Serv.IT")
  names(values.8) <- c("01010101", "01010102",
                       "01010103", "01010104",
                       "01010105", "01010106",
                       "01010107", "01010108",
                       "01010109","01010408")
  values.6 <- c("Serv.Construction","Serv.Other",
                 "Mining","Electricity")
  names(values.6) <- c("010106","010104","010102",
                       "010103")

  if (is.na(s)) {return(NA)}

  leading8 <- substr(s, 1, 8)
  attempt <- values.8[leading8]
  if (!is.na(attempt)) {return(attempt)}

  leading6 <- substr(s, 1, 6)
  attempt <- values.6[leading6]
  if (!is.na(attempt)) {return(attempt)}

  leading4 <- substr(s, 1, 4)
  if (leading4 == "0102") {return("Serv.Finance")}

  return("MISTAKE")
}

This maps each firm into one of 14 coarse categories. Here are some examples of this in action:

> cmie.14.industries("0102090000000000")
"Serv.Finance"
> cmie.14.industries("0101041502000000")
"Serv.Other" 
> cmie.14.industries("0101010601010000")
"Machinery"

So in short, the function cmie.14.industries() maps a string like "0101010601010000" into a set of 14 broad industry names such as "Machinery".

Faced with a file with roughly 48,000 firm-years, at first blush, it seems that this function has to be run 48,000 times. For a given firm, this classification could change over time, so it isn't just a matter of doing this once for each firm. Here is one simple way to do it:

badway <- function(task) {
  result <- rep("", length(task))
  for (i in 1:length(task)) {
    result[i] <- cmie.14.industries(task[i])
  }
  result
}

This is just a loop that runs over everything in the supplied vector and calls cmie.14.industries() for each element. The only concession to efficiency is that the empty vector `result' is allocated ahead of time.

This proves to be quite slow. None of the standard R vectorisation ideas offer much relief.

The key idea for obtaining a leap in performance was that while I had to run through 48,000 firm-years, the industry codes actually attain only a modest list of possibilities. This makes possible a table lookup:

goodway <- function(task) {
  possibilities <- unique(task)
  values <- rep("", length(possibilities))
  for (i in 1:length(possibilities)) {
    values[i] <- cmie.14.industries(possibilities[i])
  }
  names(values) <- possibilities
  values[task]
}

For a problem of size 1000, this works out to be 13.5 times faster:

> load("task.rda")
> length(task)
[1] 1000
> system.time(res1 <- badway(task))
   user  system elapsed 
  0.030   0.000   0.031 
> system.time(res2 <- goodway(task))
   user  system elapsed 
  0.002   0.000   0.002 

This is just a demo with a 1000-sized task. In my production situation, the performance difference is even greater, since badway() calls cmie.14.industries() 48,000 times while goodway() only calls it a few hundred times.

Ownership & governance of critical financial infrastructure

SEBI has released the Bimal Jalan committee report about the ownership and governance of critical financial infrastructure. We're going to need a similar report on the questions about entry into banking also.

Governments riding in to rescue firms

What is a government to do when a company faces a near-death situation? In almost all cases, the right answer is to let the company go under: It is not the job of a government to prevent companies from dying. Indeed, creative destruction is central to the proper functioning of capitalism. Capitalism without failure is socialism for the rich.

But sometimes, the cost-benefit ratios can look startling. Sometimes, the disruption to the economy that comes from the death of a company can be rather large. Let's look at three stories.

Three examples

GM

In July 2009, the US government chose to put $50 billion into the auto maker General Motors (GM) as part of complex rescue, which included wiping out the existing shareholders and embarking on a complex restructuring of the firm. The old GM died there.
GM got back to profitability this year. Seventeen months later, in 17 November, GM got back on its feet with an IPO which raised $23.1 billion. How impressive! See this story by Michael J. de la Merced and Bill Vlasic in the New York Times. This IPO was at $33. With this IPO, the US Treasury got down from 61% ownership to 26% ownership, so this IPO was the re-privatisation of GM. From here, if the US Treasury is able to sell its remaining 0.5 billion shares at $53 a share in the future, it will fully recoup the $50 billion that went into the rescue (ignoring time value of money).

Satyam

On 7 January 2009, Satyam announced that a lot of money was missing from their balance sheet. In the aftermath of this crisis, the government put Deepak Parekh, Kiran Karnik, Tarun Das, and three others in charge. Read this interview of Deepak Parekh with Tamal Bandyopadhyay in Mint, and this blog post by John Elliott.
The new board put the firm up for sale. It was bought by Tech Mahindra. A collapse of the firm was averted; the employees and customers largely stayed in place.

UTI

When UTI got into trouble, I was opposed to government intervention. But by and large, I think the intervention worked well. US-64 unitholders did suffer losses: half of the gap between the NAV and market value was paid by the unitholders and half by the government. And the follow-through was excellent. The staff quality that MoF was able to muster on the problem was outstanding. The UTI Act was repealed, and UTI was turned into an ordinary company. `Bad UTI' was separated out by `Good UTI'. The ownership was modified including the recent work of bringing in T. Rowe Price as a shareholder. All in all, the exchequer did well when selling off the shares in SUUTI. Privatisation hasn't yet come about, but where we are is progress.

When is it right for a government to go in?

Should the US government have gone into GM? There was a fair amount of criticism of the Obama administration for the decision. There was concern that they were doing this owing to pressure from trade unions. But the outcomes have been quite nice, so (at least ex post) it looks like a good call.

In the case of Satyam, the existing shareholders were not expropriated. It can be argued that the failure of the firm was not their fault. But by that argument, many firm failures in India in the future will justify government intervention since most public shareholders are fairly powerless when the inside shareholders have over 50% shares. In his interview, Deepak Parekh says Had it happened to a consumer finance company or a small, or even big, manufacturing company, the government would not have come out and superseded the board. The normal procedures for bankruptcy and liquidation would have taken place.. I am not sure how the future will work out.

The problem of execution capability

Satyam, GM and UTI are success stories in that the government packed a mean punch in the execution. In particular, in Satyam's case, I had simply not expected that such a nice outcome could be achieved by the government. We should really admire the teams that worked on these problems.

But can we count on such high quality execution on such problems in the future? Our success in the Satyam or UTI stories should not be generalised to the view that in the future such high quality execution will always come about.

The exit strategy

The really amazing feature of the GM story is the clarity and commitment of the government in getting out of `Government Motors' by doing a privatisation just 17 months after going in. All too often, government interventions turn into nationalisation and then you're stuck with a public sector company for a long time, with all the usual politics of the privatisation.

In the deep past, numerous weak companies have been nationalised in the decades of Indian socialism (e.g. National Textile Company) and generally the outcomes have been bad.

A particularly attractive feature of the Satyam story is that no government money was involved. The presence of government money makes things much harder. In India, all too often, it's easy to ask for government money and it's easy to get it. And if the government had got shares in Satyam, it's not easy to see how they would have got out of it.

Similarly, a nice feature of the UTI story is that in the end, the UTI Act was repealed, and UTI is on course for turning into a normal financial firm. Government intervention in the rescue did not yield an ossified PSU.

At the same time, while Satyam and UTI are good stories in terms of the exit path, we cannot generalise too much from this given the fact that GOI is at a standstill on privatisation. In general, we have to assume that what is purchased is never sold, which puts a crimp on a vast array of situations where government intervention might be evaluated.

To summarise

When most firms approach death, the decent thing to do is to let the firm die. We must rejoice in the extent to which Indian capitalism is able to bring about a steady pace of firm death. Building a good quality bankruptcy mechanism will increase the class of firms where resolution is handled in a routine and humdrum way, without the possibility of a special intervention. (Note that going through the bankruptcy process was an integral part of the GM story).

When a potential intervention situation arises, six questions need to be asked:

1. Are the negative externalities of firm death really that onerous?
2. Can government intervention be envisaged without requiring money?
3. Are the Union ministers involved in the problem known for being smart and clean?
4. Can a top quality team be put together which will work on a time-bound project starting from intervention until exit? Does this team combine competence with cleanness?
5. Do we see an exit strategy through which, within a short time, the firm will be fully out of government hands?
6. Are we very sure that in the end, we will endup imposing no costs upon the government?

Ex post, these questions worked out well for GM, UTI and Satyam.