Tuesday, August 23, 2011

More on housing and job mobility

A little while back I argued that reduced job mobility is in part attributable to the bad housing market. There's been some new evidence against certain parts of this story that are worth sharing.

Adam Ozimek shares this new JEP paper, which argues that there is no relation between the housing crisis and declines in mobility. Ozimek concedes defeat... I'm not quite as willing because I have a few lingering questions about the paper. First, what's good that the paper points out is that there has been a consistent trend for a long time now of declining mobility, and in part of course anything going on now is continuing that trend. It also provides a detailed discussion of the wide differences between CPS, ACS, and IRS estimates of mobility that should give us pause. My biggest question is exactly why "underwater" houses should be driving this. The authors look at the state-level relationship between underwater houses and mobility and find a modest positive relationship. I'm not sure exactly what this is supposed to demonstrate. First, I thought if your house was underwater you had an incentive to walk away from it, so this sub-group of particularly bad-off homeowners ought to be more mobile, right? What I was initially thinking of were people like my parents, who are not underwater but who have taken a hit on both their 401k and their home value and aren't talking about moving now like they were a couple years ago.

So the article is worth looking at, but I still have questions.

The commenter Nonymous had a lot to say on my old post on this. Most of his comments weren't convincing - they amounted to cross-sectional observations that younger people were the most likely to be unemployed but the least likely to own a home. This is Casey Mulligan stuff. Bad inferences from relative differences. My comments from before still stand on all those points. But he also shared a paper that allegedly addressed this point, which I didn't get a chance to address. It presented evidence that there were declines in mobility, but that they were mostly within-county declines, which the authors argue shouldn't affect the labor market. Anyone familiar with the spatial mismatch literature in labor economics knows that this is wrong. Labor markets in urban centers and urban peripheries are different. Counties are big. Counties are not labor markets. If you have a car, the constraints of a cross-county job are less severe of course. But think about Los Angeles County. Think about Cook County. So I'm not sure why they're presuming that declines in within-county moves don't limit job mobility.

I obviously think the biggest problem for jobs is uncertainty and deficient demand, but the impact of the crisis on peoples' assets makes a difference too - particularly when we think about immobility. This JEP paper is worth reading, and I would say it makes me think that the housing market may be less important than I had originally thought - but I still have a couple of lingering questions and I wouldn't say it's not important.

15 comments:

  1. "This is Casey Mulligan stuff. Bad inferences from relative differences."

    we need a term for this specific error.

    "never go full mulligan"

    "he did a mulligan"

    "she committed mulliganism"

    "he's a mulligator."

    ReplyDelete
  2. Hmm, why am I reminded of the work of Rene Girard?

    ReplyDelete
  3. I (coming from the humanities, not the social sciences) must be doing a bad job relating my point.

    What I'm trying to get at is:
    1) The young are historically the most mobile.
    2) The young are typically renters, not owners (so underwater mortgages should matter little to this group).
    3) The young currently have the highest unemployment rate, thus making them (historically) likely to move to find work.
    4) The young are not moving as much as they did in past economic downturns.
    5) Therefore, there seems to be something more than home prices at play in explaining reduced migration in 2007-11.

    Now, the university's proxy seems to be down, so I don't have journal access at this time, so bear with me if I don't have the right links.

    First, is the JEP paper referred to above the same as this paper: http://www.federalreserve.gov/pubs/feds/2011/201130/201130pap.pdf ?

    If so, I'll point out this from page 20 of my link: "However, both of the housing-related mechanisms proposed above suggest that migration rates should have fallen more for those who are homeowners in the current year, either because negative equity prevents homeowners from moving or because pessimistic expectations for the return on housing should not impede the mobility of renters. In both the CPS and the ACS, neither inter-state nor inter-county migration rates fell more for homeowners than they did for renters in ercentage point terms."

    So, in percentage terms, homeowners are not less mobile than renters. This is a problem if home prices are supposed to explain "much" of the immobility (or if they are an "important" factor).

    It appears that migration rates for all demographics have been and continue to be declining. This is why I asked you to consider historical migrations (Great Depression-era Okies, or Great Migration-era Black youth, or the 1890s rural flight, etc.). It appears that there is a "rise of secular rotedness." http://onlinelibrary.wiley.com/doi/10.1002/psp.670/full (Again, I can't verify that is the article I'm thinking of, but the abstract looks like what I read in June.) This phenomenon appears to have little to do with housing prices, and more to do with a general decline in migration in general, amplified by greater uncertainty, costs of moving, and other 2007-2011 factors.

    As to your distinction between in-county and out-of-county moves, it might fit this post, but your original post was addressing Tyler Cowen's OPED paragraph that I read as containing a comparison with past migrations in recessions/depressions. As the partial list above shows, those migrations were out-of state. Can you even find the discrepancies between unemployment rates within counties? (Genuine question.) The data I've seen for labor arbitrage is at the state and regional level.

    The question isn't "Why aren't people moving 25 miles to be closer to a specific job?", but rather, "Why aren't the long-term unemployed in places like Detroit (13.7% unemployment) moving to places like Nashua, NH (5.4%)?"

    There are a lot of good potential answers to the latter, underwater mortgages don't appear to explain "much" of the problem.

    ReplyDelete
  4. Nonymous -
    First, re: "Therefore, there seems to be something more than home prices at play in explaining reduced migration in 2007-11"

    I'm sure. But home prices (and really all asset prices - if your assets you might put towards a home are depleted too, that's also going to have an impact) provide a good explanation, and housing restraint have been shown to be an important factor in spatial mismatch in the past. And so far, while some of this evidence is interesting and causes me to adjust my priors a little, it doesn't really seem to challenge that.

    As for your first four points (which is the reason why I mentioned Casey Mulligan), there's a difference between noting this distinction between young and old workers at a point in time, and arguing that that relative snapshot difference is driving changes over time. Does that make sense? Mulligan does this too - he says youth are employed more in the summers so there are obviously jobs out there. If you want to say something about labor demand in the business cycle you have to take all these snapshot relative demand situations (ie - summer vs. spring employment/young vs. old employment, or young vs. old housing/ young vs. old employment) and see how they change over time. What Mulligan missed was that if you compare summer to summer you still see a decline in teen unemployment! Just like with Mulligan, I think in your case of youth homeownership something that's unrelated to the business cycle is causing youth to be (1.) more likely to be unemployed, and (2.) more likely not to be homeowners so that you can't really take the relationship between the two to be causal.

    ReplyDelete
  5. re: "So, in percentage terms, homeowners are not less mobile than renters. This is a problem if home prices are supposed to explain "much" of the immobility (or if they are an "important" factor)."

    Yes, potentially. This is why I'm starting to adjust my priors. But I'm still not totally willing to come down on the other side yet. Yes, mortgage to rent ratios have come down. But so have mortgage originations. That's not just a demand side phenomenon, Nonymous. I'm not trying to be difficult - noting that this has shifted my priors is a big deal, but it still leaves a lot unanswered.

    re: "It appears that migration rates for all demographics have been and continue to be declining."

    Yes, this was a good historical trend to point out.

    re: "This is why I asked you to consider historical migrations (Great Depression-era Okies, or Great Migration-era Black youth, or the 1890s rural flight, etc.)."

    But this is still completely irrelevant. People will move when the benefits of moving outweight the costs. Pointing to a historical case where the benefits far exceeded the costs, causing lots of people to get up and move, doesn't say anything at all baout the current situation (unless you want to argue that the spread between the benefit of moving to a new job and the cost given declining home values is as high as it was during the Great Migration, of course - if you're interested in making this argument, do it - but so far you haven't provided the evidence).

    I'm not sure I understand your in-county move paragraph. My only point is that people definitely move in-county for jobs, particularly people who rely on public transportation.

    re: "The question isn't "Why aren't people moving 25 miles to be closer to a specific job?", but rather, "Why aren't the long-term unemployed in places like Detroit (13.7% unemployment) moving to places like Nashua, NH (5.4%)?""

    Why? Economists are typically interested in decision making on the margin, since most decisions are made closer to the margin. Detroit to Nashua is not close to the margin.

    re: "There are a lot of good potential answers to the latter, underwater mortgages don't appear to explain "much" of the problem."

    I suppose I just don't see the lack of a mass exodus to Nashua, NE as "the problem".

    ReplyDelete
  6. Nonymous,

    You win the internets today.

    ReplyDelete
  7. I'm not sure what you're impressed by Gary, but you're still making me suspicious that like that other pseudonym the other day this is coming from your computer too.

    Does anyone know how to check IP addresses of commenters on blogger? Not sure if it's the case, but that would be funny.

    ReplyDelete
  8. Can you provide some evidence that housing matters a lot for explaining the reduction in job mobility. For example, how much (proportionally) has the job mobility of homeowners fallen relative to the total fall in mobility? Arguing that Nonymous' relationship may not be causal is not the same as arguing that housing values affect job mobility. At best you are saying "we don't know".

    If homeowners never moved much in the past, it kind of constrains the potential that stories which involve homeowners becoming less mobile have for understanding aggregate changes in mobility. This is all Nonymous is saying.

    ReplyDelete
  9. I don't pretend to know what proportion it explains - I couldn't say.

    People will move when job prospects are better elsewhere and the cost of moving is lower. We know that job prospects are worse than they were everywhere but they were particularly bad in places that bore the brunt of the housing crash, and we know the costs of moving - relative to before the crisis - are higher for people in places that bore the brunt of the housing crash.

    The impact on peope living in less-crashed places is somewhat indeterminate. They could move into a bad-housing-crash area more cheaply than previously, but the jobs aren't there. The impact on the people living also indeterminate - the costs of moving are higher than they were previously, but the relative payoff has increased too.

    Those indeterminancies mean its hard to pull a number out from the data. You could call that "we don't know", because we don't exactly. But I feel comfortable saying "a housing crisis is worse for mobility than other kinds of crisis".

    re: "If homeowners never moved much in the past, it kind of constrains the potential that stories which involve homeowners becoming less mobile have for understanding aggregate changes in mobility. This is all Nonymous is saying."

    But they have moved in the past, Andrew. And they still move today.



    My big concern with these articles is that I don't understand why I should care all that much about under water homes, and I don't understand why I should care all that much about within county vs. between county moves. But it's suggestive, which is exactly why I posted it.

    My big concern with most of what Nonymous has said is that he's (1.) infering from a cross sectional difference, and (2.) when he looks at changes over time he's looking at a completely different historical period with a very different set of payoffs. The only point he makes that I really like is when he mentions the article that I shared in the first place (because I thought it made good points!).

    ReplyDelete
  10. How are my first four points only snapshots?

    All of those are based on historical data as well as contemporary. Labor migration in the past has been typically a youth movement. Historically the young have had lower rates of homeownership. In the recent past through the present, the young aren't moving as much as they did in past times.

    Now, perhaps the problem is with my terms "youth" and "young," which, based on the Mulligan example you provide, you assume to mean the 16-19 year old cohort. While I'm interested (to some extent) in that group, I was really talking about the 20-24 and 25-29 groups, especially those with college degrees, but even the HS-educated show lower than average mobility. For these cohorts, the explanations I've seen for high unemployment are all business cycle ones.

    When it comes to the effects of housing/asset prices on those groups, I suggest it is minimal. If anything, the drop in home prices (if it is met with adequate credit, which is a debatable if) should encourage some movement, though I haven't seen any data that suggests that popped-bubble areas are ripe for migration. The regions with lower unemployment/higher job listings are, mostly, ones that did not have a housing bubble or its deflation.

    Credit might very well be the deciding factor here, but again this population is one that, historically is a renter population, so it wouldn't seem to be a radical cultural adjustment to rent. But, when examining the immediate post-college cohort in particular, I suspect that there is a non-business cycle explanation for renting, but for the 25-29 cohort I suspect it's quite related to the "credit crisis." Still, those two groups are historically less likely to have been homeowners, making it easier for them to historically be more mobile. Any problems of mobility related to falling home prices should be very weak in these groups. Yet, they're still less mobile than historically.

    Your point about 401ks and moving seems a bit off. If the value of the 401k affects a family's decision to move, how would they be moving for work? I mean, unless they were planning on cashing out early, paying the penalties and taxes, then wipe out their retirement savings for a new job, which I guess could be the case for the very tail of the margin.

    ReplyDelete
  11. Nonymous - you initially presented the relationship as "youth have more unemployment and less home ownership - why don't they just move?" That's a cross-sectional inference. The sort of inference we need is "given an exogenous change in housing mobility of people, what is the impact on unemployment". You can't take the fact that unemployment and low home ownership co-occur in youth (or any other age category - the age isn't the issue) as evidence that a relationship doesn't exist.

    re: "The regions with lower unemployment/higher job listings are, mostly, ones that did not have a housing bubble or its deflation"

    I agree - I make this point above.

    Sorry - I should have said savings rather than 401k. 401K plays into discussions I've had with my parents because they've been talking about retirement moves which have to be put off because they feel the need to work longer now.

    But in terms of savings in general - if the savings that Kate and I have put away took a big hit, we would have to wait longer to buy a house than we originally planned. But you're right, our 401k specifically is beside the point.

    ReplyDelete
  12. "you initially presented the relationship as "youth have more unemployment and less home ownership - why don't they just move?""

    According to Google, I did not such thing. My original use of age as a demographic consideration to consider was this: "If homeownership is directly related to age, and unemployment rates are inversely related to age, then how can "much" of the low mobility among the unemployed (young) be explained by housing (old)?"

    Which, admittedly is quite clumsy, so I'll try again:

    You assert that "much of this [low labor mobility] is due to housing conditions."

    I tried to take housing out of the question of labor mobility by pointing to a demographic that is not now, nor historically has been predominantly homeowners (i.e. 20-29 year olds). I then pointed out, least you assume otherwise, that that group has (and has had) a higher rate of unemployment that all other demographics except for teens (I should have been more specific). I then pointed out that the group historically was the most mobile in terms of job search (and the data is for extra-county moves as provided by the ACS and census). Finally, I pointed out that the mobility of that group (20-29 year olds) was also in decline. All that was to argue that, even when you have a group with high unemployment, and you (largely) remove current homeownership from the equation, labor mobility is still down.

    ReplyDelete
  13. "Why? Economists are typically interested in decision making on the margin, since most decisions are made closer to the margin. Detroit to Nashua is not close to the margin."

    Maybe because that's the sort of mobility that Cowen is concerned about? Sure, that's not clear in that small passage alone, but since he's talking about historical numbers, and since we don't really have those for in-county moves, but we do for metro-to-metro and interstate moves that was what I got.

    "People will move when the benefits of moving outweight the costs. Pointing to a historical case where the benefits far exceeded the costs, causing lots of people to get up and move, doesn't say anything at all baout the current situation"

    I agree. For the examples I provided, most homes weren't sold but abandoned. People moved to areas where they predominantly rented, not immediately bought (but, that's historically the case regardless). The costs of moving appear to be higher than the benefits. Some of that, for some people might be their home's value/price. But, the literature doesn't support that very strongly. In fact, most of it is struggling to find the true cost, which I suggest is a combination of cultural factors as much as the economic ones. This is the crux of the "rise of secular rootedness." It might also be part of an explanation for why so many college grads have moved back in with their parents (even before the housing crash).

    As to identifying IP addresses, you will evidently need a stat counter: http://www.google.com/support/forum/p/blogger/thread?tid=1c9974c8efe48af9&hl=en

    ReplyDelete
  14. Daniel, is there an e-mail address I could reach you at? Or are the blog threads the best ways to interact? I have some questions that are slightly off-topic.

    ReplyDelete
  15. dan dot p dot kuehn at gmail.com

    Off topic questions are fine here too.

    ReplyDelete

All anonymous comments will be deleted. Consistent pseudonyms are fine.