<b>Rohit Lamba & Arvind Subramanian:</b> Why India should focus on leverage not just capital adequacy

In part five of a series, the authors explain that if CRAR is proving to be tricky internationally, exclusive reliance on it in India is perhaps even more fraught with risks

Should India change the way it assesses the stability of the banks? The answer may be yes.

One of the consequences of the Great Recession in the West (2008-13) has been an active search for better measures of banking stability. The standard measure - and one that is still preferred by national and international regulators, including Basel-III - is the ratio of capital to risk-weighted assets. In India, too, this avatar, called CRAR - capital to risk (weighted) assets ratio - has been the dominant measure for assessing bank stability.

According to the Reserve Bank of India (RBI), CRAR is calculated by "dividing the capital of the bank with aggregated risk weighted assets for credit risk, market risk and operational risk." Intuitively, the measure makes sense because it captures how well-cushioned or capitalised banks are relative to the riskiness of their assets.

But the Great Recession exposed an almost fatal flaw in the CRAR. Leading up to the Lehman crisis, the CRAR did not show any serious deterioration, it did not flash amber. For example, 15 days before its collapse, Lehman Brothers boasted a Basel-type capital adequacy of 11 per cent, suggesting strength and stability not warning about near-bankruptcy. The reason, of course, is that assessment of risks is inherently tricky, more dark art than science, and hence prone to being misleading, especially when left to the bank itself. As a wag noted, self-assessment is to assessment what self-praise is to praise.

Unsurprisingly, therefore, the international focus is shifting in giving more weight to a much simpler measure of banking stability, namely the leverage ratio (LR), which is defined as the ratio of total capital to total assets. Conceptually, to make it foolproof, only Tier-I capital, typically equity, is to be included in the numerator, and all assets - on and off the balance sheet - form the denominator. Although simpler, indeed primitive compared to CRAR, there is now enough evidence to show that the LR would have better predicted the stress in the banking system than CRAR in the run-up to the financial crisis.

In principle, if risk weights predicted the underlying uncertainty accurately, the two measures - CRAR and LR - should be highly correlated. Before the onset of the global financial crises, this correlation weakened and eventually turned negative.

Figure 1 plots the time series of the correlation between the two indicators for Europe and India. In Europe, the correlation steadily declined over the last decade, with alarmingly negative numbers since 2007. For the public sector banks (PSBs) in India, the correlation between CRAR and LR has picked up again in the last few years, after having worryingly dipped to less than 0.1 in 2010.

While Figure 1 plots the correlation over time, Figure 2 shows a scatter plot for the last three years' average of CRAR and LR for all PSBs in India. The trend-line is positively sloped, which is good news. However, the correlation, at 0.45, is not particularly strong, especially when viewed in the context of rising toxic assets. Most importantly, there are some worrying outliers (south-west corner) in Figure 2 that merit examination. For them, the CRAR is misleading: it suggests that banks are adequately capitalised when they, in fact, are over-leveraged and hence vulnerable.

Beyond correlation, it is also important to ask what is a safe level of the LR. For Indian banks, the leverage ratio for PSBs varies from 7.8 to 4.5, which is well above the Basel-III benchmark of 3 per cent. But is 3 per cent safe? In their new book, Bankers' New Clothes, Anat Admati and Martin Hellwig argue that a 3 per cent leverage ratio for a bank implies that a bank will go bankrupt if its assets lose more than 3 per cent in value. Importantly, banks themselves would never grant loans to a firm that only had only 3 per cent effective equity. The authors propose a much higher leverage ratio in excess of 10 per cent, even 15 per cent. Some critics, however, argue that high leverage will significantly reduce profitability because it will curtail the amount that a bank can lend and invest.

Against the 10 per cent benchmark, all the PSBs fall short but most private sector banks don't as Figure 3 suggests. It plots the LR across time for public and private sector banks in India, as well as the situation of the second-worst and second-best PSBs. The figure highlights that there is substantial variation in the LR within PSBs as well as between public and private sector banks, making it a sensible measure of differentiation among banks. Strikingly, even the best public sector banks are much below the average private sector bank.

If CRAR is proving to be tricky internationally, exclusive reliance on it in India is perhaps even more fraught with risks. Given the relatively weak governance systems within our banks, and the difficulty of regulating them from the outside, there is even more reason to be anxious about current risk assessments.

Adding to the problem today is the large amount of stressed assets in banks. How are these being assessed and classified? There may be a strong incentive to "extend and pretend," classifying stressed assets as better than they will actually turn out to be. Under-recognising bad assets is, after all, an age-old problem. In these circumstances, the CRAR measure is probably highly distorted, and the LR measure simpler, more transparent and hence more useful.

The leverage ratio should, therefore, occupy a greater space in regulatory discourse in India.

Lamba is a post-doctoral fellow at Cambridge-INET, Faculty of Economics, University of Cambridge, and Subramanian is chief economic advisor in the Ministry of Finance
Part four appeared last Friday