For some decades, governments worldwide were quite infatuated with ambitious guaranteed pension programmes. It was felt that a government made people very happy by guaranteeing pension in old age, and that this was something that was feasible.

From roughly 1980 onwards, it has become increasingly clear that the task of giving out guaranteed pension is fraught with danger. The essence of the argument is this. When a young man joins the labour force at 20, his pension at 80 depends on interest rates, asset market outcomes, and mortality changes over the next 60 years. What will interest rates be like? Will new advances in biotechnology dramatically extend the lifespan? Who knows anything about what will happen to these three aspects of the economy in the next 60 years?

Many countries, particularly in Europe, made big mistakes in setting up grandiose "defined-benefit" pension systems, which made promises to all citizens. At the time, this seemed okay, for there were only a few old people and a lot of young people. But who could have predicted what happened in the next few decades?

In the event, what happened in Europe was a dramatic decline in birth rates, and an increase in the number of old people. This has given rise to a situation where defined-benefit pension has become a major source of macroeconomic distress. The European welfare state is dying, and mistakes in pension economics are one key part of the causes of its demise.

In India, in 1995, one important policy blunder took place, in the creation of the Employee Pension Scheme by the EPFO. This is a defined- benefit system. Economists in 1995 knew that this scheme was a mistake, but it was supported by trade unions and left parties. Today the scheme has a hole of roughly Rs 20,000 crore (Rs 200 billion), and the hole is rapidly growing in every month of inaction.

In India's case, we know for sure that in the years to come, birth rates will drop and there will be an increase in the number of old people. Hence, even though setting up a defined-benefit pension system might be politically attractive, it is surely wrong.

Hence, the New Pension System that has been adopted by the Government of India and by most large states is a pure defined-contribution system. The young man joining the labour force at 20 is given a systematic planning framework, where he saves money every month.

The money is invested in an NAV-based system. There is no guarantee given by the government about what his pension at 80 will be. We know that he will do pretty well, but there is no attempt at writing down a number (or a minimum level) specifying what he will get.

Now there are calls for pension guarantees. As with defined-benefit pension, pension guarantees are politically seductive and potentially very dangerous. Many alternative guarantee structures can be thought up, and to a layman, they sound like reasonable propositions.

They often involve extremely high costs. These issues were analysed in an EPW article in 2003 titled "Investment risk in the Indian pension sector and the role for pension guarantees" (see http://tinyurl.com/ygjfys for a paper and free software). The main finding of the paper is that many plausible-looking guarantee structures have extremely unpredictable distributional effects and can involve very high prices.

To fix ideas, let us consider a person who earns Rs 3,000 per month at 20. We analyse an NPS-style system with a contribution rate of only 8.33 per cent. The results are, hence, not directly comparable with the NPS, which has a contribution rate of 20 per cent. However, the qualitative features of the results will hold for the NPS also.

Suppose a guarantee is made that the pension will exceed Rs 50 per day, i.e. that of comfortably going past the poverty line. Such a guarantee is worthless if investment is only in government bonds, because he is going to beat poverty anyway.

But if the person invests 100 per cent in equities, then the guarantee is useful, because there is a finite probability of going into poverty. The cost of the guarantee in that case is Rs 14,000 per person. So this guarantee helps people who invest more in equity.

The guarantee that trade unions like most is the one that pension will exceed 50 per cent of the last wage. This guarantee is very expensive. If the person has no equity, it costs Rs 61,000 per person. If the person is 100 per cent invested in equity, the guarantee costs Rs 70,000 per person. This translates into very big numbers when we think on the scale of 10 million or more civil servants. If the government adopts a liability of Rs 60,000 per worker for 10 million workers, it is an implicit debt of Rs 60,000 crore (Rs 600 billion).

The third guarantee structure that comes to mind is a "capital protection guarantee" which guarantees real returns of at least zero per cent. For a person holding only government bonds, this guarantee is worthless. For a person with 100 per cent invested in equity, it costs Rs 4,500 per person. So this guarantee helps people who invest more in equity.

These results vary in subtle ways with every small detail of the pension system and the framework used for the calculations. Many surprising results come up in these calculations. Plausible-sounding guarantees can work out to be very expensive, and the identity of the beneficiaries is often surprising. Trade union leaders will probably not be excited that a guarantee to beat poverty is most useful for people who buy 100 per cent equities.

Guarantees are not free. The best path is for the government to give out no guarantees at all, and leave it to the choice of workers.

Some workers will chose to invest in government bonds and take no investment risk. Some workers will choose to hold 50 per cent in equities. Some workers will choose to have 100 per cent in equities and buy guarantees (thus gaining some safety but getting lower returns).