Unbelievable balance sheets: an out-of

RCEA Macro-Money-Finance Workshop 2016
Unbelievable balance sheets: an out-of-sample procedure
for forecasting the negative net worth in banking
Mikhail Mamonov
This paper proposes an empirical framework for estimating the scope of negative net worth on
banks’ balance sheets. Banks operate with negative net worth, or hidden negative capital, in case
they have too aggressive lending strategies to create sufficient amount of loan loss provisions
(LLP). To satisfy the official requirements on capital adequacy such banks artificially increase
their capital, but eventually they face with license withdrawal. Using the latest bank-level data on
Russian banks, we show that this story really happens and attracts growing concern from
monetary authorities. Russian banking system can be treated as a valid laboratory for testing
these effects since in the last few years there was a surge in license withdrawals caused by the
Central Bank of Russia’s intention to clean the system from unfair competitors. First, collecting
the data from various issues of the Bulletin of the Central Bank of Russia over the 2014-2016, we
show that the revealed size of hidden negative capital in a sample of 106 credit institutions,
whose licenses were withdrawn during this period, amounts to about RUB 1450 bln (2% of
Russia's GDP in 2015). Second, following the literature on the cost of banking failure developed
on the basis of US banking by James (1991), Schaeck (2008), Bennet and Unal (2014), Cole and
White (2015) and others, we first estimate a set of simple OLS-models showing that some factors
affecting the cost of failure in Russia are very similar to that in US (bank size, capital, LLP,
excessive asset growth). Next, we introduce a set of new hypotheses regarding the reasons of
negative net worth accumulation on balance sheets, namely the high turnover and the low
turnover hypotheses, and empirically test them within the OLS-models. Third, based on the
previous step, we incorporate the determinants of negative net worth into a Tobit model built on
the sample of 106 failed banks and the subsample of those banks that a-priori have minor
incentives to falsify their balance sheets (largest government-owned banks and foreign banks).
Using this Tobit model, we produce the out-of-sample forecasts for the other part of Russian
banking system. Our estimations have shown that about 250 of still operating credit institutions
in Russia falsify their balance sheets so far and the estimated out-of-sample forecast of their
cumulative negative capital amounts to RUB 900 bln. Importantly, if these banks leave the credit
market, the non-financial firms and households will lose approximately RUB 4500 bln of credit
resources. To deal with possible sample selection bias in Tobit model, we check the robustness
of our results by estimating Heckman selection model. We suppose these results might be useful
for monetary authorities and contribute to the literature on banking stability.