Financial bubbles burst with varying effects depending on the kinds of assets that rose in value in the boom years. Macroeconomists, financial analysts, and policymakers alike are aware of these differences particularly when one compares the disastrous housing market bubble-and-burst that sparked the Great Recession of 2007-2009 compared to the less damaging 2001 “dot-com” stock market recession. But what exactly distinguishes the length and severity of a post-bubble recession?

New research by Oscar Jorda of the Federal Reserve Bank of San Francisco, Moritz Schularick of the University of Bonn, and Alan M. Taylor of the University of California-Davis looks at historical data on economic growth, credit growth, and financial assets to understand the effects of bubbles. Their data set covers 17 countries since the 1870s, including the years directly after the Great Recession, among them the United Kingdom, Germany, France, the United States, and Japan.

They find, unsurprisingly, that the bursting of asset bubbles lead to weaker recoveries on average compared to typical down swings in the business cycle. But there are substantial differences in the effects of different kinds of bubbles when they collapse. Not all bubbles inflate economies in the same way, and when they burst economies react differently.

Jorda, Schularick, and Taylor split bubbles into four groups, divided along two lines. The first broad way to categorize bubbles is by the kind of asset the bubble inflates. A bubble might be in equities, such as the high-tech stock bubble in the United States during the late 1990s. Or the bubble could be in housing, which of course took down the U.S. economy and then other leading economies around the world only eight years ago. The second way to categorize bubbles is to see if they happened concurrently with large increases in credit. In other words, were the bubbles financed with debt? So we end up with four kinds of bubbles: equity bubbles without credit bubbles, credit-fueled equity bubbles, housing bubbles with average credit growth, and leveraged housing bubbles.

The three economists find a hierarchy for the effects of bubbles. Bubbles in equity assets that aren’t financed by credit aren’t particularly virulent. In fact, Jorda, Schularcik, and Taylor find that these bubbles don’t make recessions any worse. Or at least, there is no statistical difference. Debt-fueled equity bubbles are more damaging, making recessions more severe and subsequent economic recoveries slower. Yet housing bubbles are even more damaging. Even in the absence of a large credit build up, housing bubbles are quite harmful to the broader economy.

But when mixed together with credit, leveraged housing bubbles become extremely powerful. The economists find that economies that experience these kinds of bubbles don’t fully recover from the recession until, on average, five years after the bursting of the bubble. Indeed, consider how weak the current U.S. recovery has been or the long depression in Japan after the collapse of a real estate bubble there. These results have obvious implications for macroeconomic policy, particularly when it comes to monetary policy. We should be on the look out for credit-fueled housing bubbles.

But there’s another interesting angle to these results. As economists Atif Mian at Princeton University and Amir Sufi at the University of Chicago point out in their recent research, the distribution of credit is very much about the distribution of wealth. Wealthy households who have more assets and a positive net worth put their savings into financial institutions that then increase credit to borrowers, most likely to more and more low-income individuals as debt-driven bubbles inflate. But debt is the opposite of insurance, making the overall economic system more fragile when too much debt builds up among those least able to repay it when economies over-inflate.

At the same time, assets are owned by very different segments of the population. Equity ownership is much more concentrated in the hands of the rich, while housing is more broadly owned. A bursting of a bubble in an asset more widely held will have a stronger impact on the economy. Mian and Sufi point out this difference as a key reason for why the recession following the bursting of the tech stock was less damaging than the 2002-2006 housing bubble.

This isn’t to say that wealth inequality is always at the heart of financial bubbles. But it appears that it could play an important role either in the inflation of asset bubbles or in determining the level of damages once they burst. Policymakers need to understand these differences and the underlying factors behind them when it comes to housing policy, credit policies more broadly, financial market supervision, and if we think wealth inequality plays a role, tax policy.