Wednesday, June 2, 2010

On the cost of financial crises

Are financial crises costly? To answer this question, one should not look at the cost of a bailout, a drop in GDP or missing tax revenue, but at what people care about: consumption. In this regard, the current crisis is too young to be analyzed, but other ones are available. Two recent papers look at this for Japan and Norway.

Yasuyuki Sawada, Kazumitsu Nawata, Masako Ii and Mark Lee use panel data from Japan that spans over the 1997 banking crisis and estimate Euler equation that allow for credit constraints. While in normal times, 7.82% of households are credit constraint, this increases only to 8.44% during the credit crunch. In other words, the ability for households to smooth out consumption was only negligibly affected.

Eilev Jansen studies Norway, but prefers a VAR approach linking current wealth and income to consumption, which appears to work better than Euler equation approaches for the recent years. But again, the impact of the crisis on consumption is negligible: the elasticity of equity income on consumption is 2%.

Thus, the impact on consumption seems to be minimal. So why again are we seeing these huge interventions?


Anonymous said...

But that's data from countries where we had interventions, so these are the costs of financial crises to which there was government involvement and expectations of policy involvement. I'm not sure what's the counterfactual path of consumption we would need to consider.

It's very hard to convince politicians to "do nothing" for the sake of scientific progress.

Eilev Jansen said...

Thanks for your interest in the paper.
However, the paper is not correctly cited . The long run elasticity of wealth (housing + net financial wealth) is estimated to 15 per cent. So the impact is in fact large!
The LR elasticity of equity income ( of 2 percent) cited is from a previuos study which is encomappsed by the new consumption fn. Please see link to paper.

Eilev S. Jansen
Statistics Norway

Economic Logician said...

Eilev: that is the problem with empirical papers presenting many different specifications. My take was that the "definitive regression" was equation (4), which shows 2%. I agree, other specifications show 15-20%, but for short-term effects only.