Against the advice of many economists, emerging markets and developing countries increased reserves massively since the late 90s (See here for instance). While the increase in Latin American countries took predominantly place in the mid 90s, the rush in other nations followed the Asian crisis.
One of the most prominent of those has other worries since last month. On the other end of the spectrum, others have already claimed that reserve levels have not been high enough Arvind Subramanian concludes in a recent VOX article that “Lesson number two [from the financial crisis] is that self-insurance helps and absolute self-insurance helps absolutely.“ Unlike, Summers who advised the Indian central bank not long ago to diversify its reserves and claimed its excess reserves, to amount to 15% of GDP in 2006, Subramanian now advises a level in the future of US 1 trillion! Which corresponds to 85% of the current GDP! Currently reserves stand at 21% of GDP roughly down by 18% (or 5 percentage points) compared to their peak a quarter ago. Hence, Subramanian advices a level roughly four times higher then current reserves while Summers was favorable to a level of a half of the current level.
Despite all the criticism to Lawrence Summers findings, one of his points remains strong and seems under weighted by Subramanian: the cost. A rough guesstimate of the cost of holding the reserve level is given by the difference between the interest on domestic investment and the return on the reserves times the level of reserves. Let the return on reserves be around 3 percent and the return on domestic investment be about 6%. The gap is most likely a lower bound rather than an upper bound. The yearly implied cost of holding reserves equivalent to 85% of GDP is roughly 2.5% of GDP! Is the insurance worth it to forgo 2.5% of growth each year? It seems outrageous given that we are generally happy to find policies which enhance growth by much less.
A structured way to think about reserve levels is based on some notion of insurance against an expected loss with a certain probability. But no optimizing agent (irrespective of the level risk aversion) would want to sacrifice so much money unless there is an extraordinary high chance of a loss in the subsequent period or the loss is expected to amount to a level well above what we have seen and will even see in this crisis.
So why such a high level? Three reasons remain: 1) Actual costs are much lower (i.e. 0.03 is just wrong and much closer to zero), 2) the authority wants to defend an exchange rate level or 3) the authority is just not minimizing in a rational manner at the margin.
Option 1) and 3), though possible, are rather awkward. Option 2) however remains pretty strong. But then again sticking to a certain level of the nominal exchange rate needs to deliver substantial gains in terms of higher GDP to justify so high levels of reserves. Gains that have not been found in empirics.
Some months ago it might have seemed that the bags are full and the central banks stand ready. Now we are asking are they full enough or the central banks / governments unwilling to make use of their hard currency? Are we going to have a new wave of countries that will move to more flexible exchange rate regimes since they are not anymore willing to pay the cost?With the advent of the financial crisis the discussion around the level of reserves has started to vanish. But it is exactly now that we need to take a stance since after the last “global” crisis in form of the Asian crisis reserve accumulation expanded strongly. It is important to clarify whether it has been a sensitive policy to increase reserves or whether the costs do not satisfy the increased levels.
 Not surprisingly for Subramanian, lesson number one read, “that the greater the financial integration, the greater the susceptibility to financial crises, and the larger their final cost.”