The most recent US downturn was so painful because US households’ borrowing binge in the first half of the 2000s left them stuck repaying large debts (often against assets that had plunged in value) and unable to spend money on new goods and services that they actually wanted. Moreover, they weren’t in a position to take out new to debt to support consumption as they had before the crisis.
A fascinating new paper by Xavier Giroud and Holger Mueller argues that many US companies went through a similar experience, and that this made the downturn about twice as worse as it otherwise would have been.
Before we dig into the new research, a quick summary of what we know about what happened household spending and employment during the boom and bust years.
American house prices rose by about 70 per cent between the start of 2002 and the end of 2006. Borrowers monetised a large portion of those gains by taking out home equity loans, by refinancing existing mortgages into new ones, and by getting capital gains when they sold their homes.
During the peak of the bubble, Alan Greenspan and James Kennedy, a senior economist at the Fed, wrote a pair of papers explaining how to measure this home equity extraction and estimating where that extra money went. Kennedy has updated these estimates through the end of 2008, available here.
The chart below shows Greenspan’s and Kennedy’s estimate of the share of household spending on home improvements and personal consumption that was directly financed by home equity withdrawals:
Relative to the 1991-2001 average, US households spent an extra 2.2 per cent on home improvements and personal consumption — nearly $1 trillion, split about half and half between the two categories — from the start of 2002 through the end of 2006. (Technically, “improvements to residential structures” are counted as investment by the Bureau of Economic Analysis, but it isn’t obvious to us that buying a new refrigerator or television is fundamentally different from installing a hot tub, so we lumped them together.)
The Greenspan and Kennedy estimates of the direct impact of the housing bubble on personal consumption almost exactly match what Atif Mian and Amir Sufi estimated more recently using a bottom-up methodology. Both suggest that America’s GDP would have been about 1.5 per cent smaller at the end of 2006 in the absence of the bubble-supported spending on consumption and home improvement, and assuming that most other things in the US economy had unfolded in roughly the same way. (Big assumptions, but useful to get a sense of the magnitude of these numbers.)
Mian and Sufi have also looked at how the growth in household debt during the boom exacerbated the downturn once house prices started falling, both by using granular data on auto purchases and credit card spending as well as changes in different kinds of employment at the local level. They repeatedly found that the places (either postal codes or counties, depending on which measure they used) where households were the most leveraged suffered far more than the places with the least-indebted households.
When it came to measuring the impact of debt on employment, Mian and Sufi cleverly divided jobs into tradeable and non-tradeable ones, on the reasonable theory that cashiers, doctors, and construction workers are more sensitive to local economic conditions than people who work at call centres or auto assembly plants. After comparing changes in local non-tradeable employment against changes in local house prices, they concluded that “a 9.5% reduction in housing net worth (which is what the economy experienced between 2007 and 2009) leads to a reduction in overall employment of 2.9%, or 55% of the actual decline in total employment of 5.3%” The housing bust is what made the recent downturn more severe than the early 1990s and early 2000s recessions.
Giroud’s and Mueller’s goal is to provide an explanation of how the collapse in house prices actually led to job losses. After all, workers don’t sack themselves. Their argument, which we will explain below, is that “all of the job losses associated with falling house prices during the Great Recession are concentrated among establishments” of firms that increased their leverage during the boom.
If you combine this finding with the estimate from Mian and Sufi, you get the conclusion that rising business leverage in 2002-2006 made the downturn about twice as painful as it otherwise would have been. Or, put another way, the housing bust would have been far less painful if companies hadn’t borrowed so much during the go-go years.
Giroud and Mueller looked at 2,800 businesses, excluding financial firms and utilities, which, as of 2006, operated 284,000 “establishments” (think of a chain restaurant with multiple locations) and employed about 11.2 million people.
We compared this sample to the universe of firms covered by the latest Census data and found that it is somewhat skewed toward larger businesses.
The average firm in the Giroud and Mueller sample had 101 establishments and 4,005 employees as of 2006. As of 2012, there were about 102,000 American companies with at least 100 employees. These firms operated a little more than 1.5 million establishments between them, employing about 76 million people (two thirds of American workers), for an average of around 15 establishments per firm and 749 workers per business. Even among the 18,219 firms that had at least 500 employees, the average company only had around 66 establishments employing 3,286 workers.
We tend to think these differences probably don’t affect the overall conclusions of the paper, but it’s something to bear in mind.
Anyhow. Giroud and Mueller sorted all these firms by the changes in their debt-to-asset ratios, based on book value, between 2002 and 2006. In the aggregate, the indebtedness of US non-financial businesses has been basically constant since 2000. But that aggregate masks very large changes within the broader sample: “about half of all establishments belong to firms that come into the Great Recession having just tightened their debt capacity, while the other half belong to firms that come into the Great Recession having just freed up debt capacity” by reducing their leverage.
That’s why the dotted line in the chart below (from their paper) is so misleading:
As of 2006, the businesses that levered up since 2002 tended to be somewhat smaller than their more conservative brethren in terms of the number of establishments they operated (13 per cent), the number of people they employed (26 per cent), and the book value of their assets (5 per cent). During the boom, however, they grew more rapidly on all three dimensions.
The debt gluttons were less profitable (around 0.6 per cent vs 4.4 per cent) and had worse returns on assets (2.6 per cent vs 6.6 per cent), perhaps because they paid their workers about 6 per cent more, on average. They also had much higher debt-to-asset ratios (38.3 per cent vs 19.5 per cent).
You might reasonably think that job losses were worse among the firms that levered up during the boom because they grew too much and because they were less efficient users of capital. Giroud and Mueller, however, believe the real problem is that their high debt levels going into the recession prevented them from borrowing to sustain employment during the downturn. Their theory is that firms that were less indebted in 2006 were willing and able to borrow to sustain employment, while the highly-levered companies had no choice but to cut jobs.
Giroud and Mueller start by using the insight from Mian and Sufi that businesses in non-tradeable industries — retail, dining, healthcare, construction, etc — adjust their workforce in response to changes in local house prices, while other businesses don’t. Unlike Mian and Sufi, they focus on the financial characteristics of the specific businesses that were responsible for the changes in non-tradeable employment during the housing bust.
There isn’t any meaningful difference in the geographical distribution of firms that levered up during the boom and those that cut their debt, which means that the two groups of companies experienced basically the same changes in house prices. As Giroud and Mueller put it, “whether firms tightened or freed up debt capacity in the run-up to the Great Recession is orthogonal to both the incidence and magnitude of household demand shocks during the Great Recession.”
That’s useful, since it means there are many places where Giroud and Mueller can compare lots of different establishments in the same industry that experienced the same exact local housing conditions, but whose parent companies had very different debt profiles. As they put it, “our specification forces comparison to be made between relatively small, local establishments in the same industry and ZIP code, e.g., a local Macy’s versus Nordstrom department store, a local Safeway versus Kroger supermarket, or a local Target versus Kmart discount retailer, where some establishments belong to high-leverage firms and others belong to low-leverage firms.”
After running the regressions, Giroud and Mueller find, unsurprisingly, that variations in local house price changes don’t affect employment at businesses in the tradeable sector. However, when they focus on non-tradeable industries, they find that all of the job losses associated with local house price conditions can be attributed to establishments of businesses that levered up during the boom. Moreover, to the extent that non-tradeable businesses shuttered establishments between 2007 and 2009 as a result of falling house prices, “the effect is entirely concentrated among establishments of high-leverage firms.”
The final step in the puzzle is to compare what the highly-indebted and less-indebted businesses did from 2007 through 2009:
High-leverage firms indeed act like financially constrained firms in the Great Recession. When faced with household demand shocks, these firms do not (or cannot) raise additional external finance. Instead, high-leverage firms reduce employment, close down establishments, and cut back on investment. The opposite is true of low-leverage firms. When faced with household demand shocks, these firms do not reduce employment, close down establishments, or cut back on investment. Instead, low-leverage firms increase both their short- and long-term borrowing, consistent with these firms having freed up debt capacity in the run-up.
Moreover, the indebted businesses tried to spread the pain of house price declines in certain localities by cutting employment in other establishments. That doesn’t happen with the less indebted businesses, presumably because they aren’t facing the same set of financial constraints.
You may recall above that the more-indebted companies were slightly less productive than their less-indebted peers and also grew a bit more, on average, during the boom. In other words, it’s possible that the way indebted businesses in non-tradeable industries respond to local house price collapses is a symptom of something other than their high debt levels.
Giroud and Mueller first tested this by sorting firms by how much they grew, both in terms of employment and in assets. If excessive expansion during the bubble were the real problem, then fast-growing companies that had cut their indebtedness would have cut many more jobs than their slowly-growing peers in response to changes in local house prices, while slowly-growing but highly-leveraged firms should have cut far fewer jobs than their fast-growing but highly-indebted peers.
That isn’t what happened. Once you separate out the changes in indebtedness, differences in pre-crisis growth rates don’t predict a given businesses’s sensitivity to changes in local house prices. Similarly, unproductive businesses that had cut their leverage didn’t respond any differently to changes in local house prices than highly productive businesses that also cut their leverage. However, high levels of productivity wasn’t helpful to businesses that boosted their indebtedness during the boom. The debt was what determined which business sacked workers when house prices fell, not the fundamentals.
None of this would have mattered if the companies that had cut their debt during the boom snatched up the workers getting laid off by the financially reckless companies during the bust. Unfortunately, this didn’t happen, despite the fact that borrowing costs fell significantly in real terms. Those who had the misfortune of living in places with a relatively high concentration of debt-engorged businesses were more likely to lose their job than people who lived elsewhere, even after controlling for changes in house prices.
The dangers of excessive household debt, especially when debt ratios soar rapidly, are relatively well-known. Now there is increasing evidence that sudden upticks in corporate debt can be even more damaging. (That said, it’s dispiriting to imagine how slowly the US economy would have grown in the first half of the 2000s in the absence of the housing bubble and excessive corporate borrowing.) Something to bear in mind given what’s been happening in China over the past few years…
Related reading:
Explaining the pain in Spain — FT Alphaville
Fix housing finance, fix the economy? — FT Alphaville
Copyright The Financial Times Limited . All rights reserved. Please don't copy articles from FT.com and redistribute by email or post to the web.