Wednesday, March 31, 2010
I have recently softened my view that the Federal Reserve gets little blame/credit for the housing boom. But shouldn't low interest rates make a building boom for all three types of structures? What did the Federal Reserve do to create a bias against non-residential building?
Added: Milton Recht (thank you!) pointed me to these three graphs. By one of the measures, non-residential vacancy rates were high during the housing boom. Is that a coincidence? In theory, low non-res construction activity could encourage housing construction, but it would do so by lowering housing prices rather than increasing them. Moreover, it seems that the purchase price of non-residential real estate was high (despite the perhaps high vacancy rates).
Government policy continues to rigidly classify mortgages by their “affordability,” and pays too little attention to strategic defaults. Policy mistakes like these serve to prolong the foreclosure crisis.
The home foreclosure process begins when a homeowner stops making his mortgage payments, so the major policy question is how to get homeowners to pay.
One school of thought is that mortgage payments are too high to be “affordable,” and if the government could arrange for payments that were a more reasonable fraction of the borrower’s income, homeowners would make their payments and keep their homes. As I explained last week, this idea is the cornerstone of federal mortgage modification programs, under both the Bush and the Obama administrations.
The other school of thought is that “affordability” cannot be so rigidly defined, and that the vast majority of homeowners — even many of the unemployed — would make their mortgage payments if they had enough incentives to do so. In this view, most foreclosures are cases of “strategic default”: homeowners who determined that it is worth a move, and a reduced credit rating, to erase their negative home equity and relieve the stress associated with sizable housing payments.
One of the criticisms in last week’s report by the Troubled Asset Relief Program’s inspector general is that federal mortgage modification programs do not reduce the principal balance of underwater mortgages, and thereby fail to address strategic defaults.
The report also pointed out that a large number of homeowners who had their mortgage payments reduced to levels that the government deemed “affordable” still defaulted. This is further evidence that the government policy will not significantly impact the foreclosure crisis by focusing on “affordability,” yet affordability is the centerpiece of the Obama administration’s new set of policy initiatives to keep unemployed workers in their homes.
As shown by a 1999 study by Martha Olney, a professor at the University of California, Berkeley, the Great Depression of the 1930s offers vivid evidence that borrowers default for strategic reasons, and make full payments even on loans that appear “unaffordable” when given the incentive to do so. (Thanks to Song Han at the Federal Reserve Board for pointing me to this study).
During the 1920s boom that preceded the Great Depression, consumer indebtedness grew twice as much as household incomes, which themselves grew significantly. From 1929 to 1933, employment and household income fell many times more than they have fallen in the current recession: If there ever was a time when households could not afford to pay their debts, the Great Depression was such a time.
Indeed, real estate mortgage defaults were common in the Great Depression, and both affordability and strategic default might be cited as reasons. But Professor Olney showed that the incentives for strategic default varied across loan types during the Great Depression, and that default on consumer installment loans made little sense from a strategic point of view.
In the 1920s, it was common for families to purchase automobiles, refrigerators, stoves, and other consumer durables with “installment debt”: that is, they made a down payment of about one-third, took the item home immediately, and promised to make regular payments until the item was fully paid. If the borrower failed to make his payments — even the last one — the item he purchased could be repossessed and he would receive no refund of his prior payments.
In other words, absent the destruction of the item purchased, it was impossible to be “underwater” on the typical installment debt contract of those days, and thus there was no incentive to strategically default.
Professor Olney found that defaults on such contracts were rare in the early 1930s: “Despite the layoffs, the wage cuts, and the unprecedented prevalence of installment credit use, families with installment debt were avoiding default” (pp. 321-2).
Later in the Great Depression, the rules for repossessing consumer durables changed, and consumers had to be given a refund of part of the payments they made prior to default. The incentive to repay installment debt fell, and Professor Olney found delinquencies and defaults on installment debt to rise.
Many people who lost their jobs in the 1930s still made their debt payments, as long as they had an incentive to do so. Today homeowners with negative equity have little financial incentive to make their payments. By focusing so much on “affordability,” the Obama adminstration’s latest policies do little to prevent strategic default, and should not be expected to alleviate the foreclosure crisis.
Tuesday, March 30, 2010
With at least 6 months more data in hand, and housing prices still seeming to fall sharply, Professor Shiller wrote in the New York Times that the housing price "decline might well continue for a long time" -- apparently disagreeing with me that a housing recovery was beginning.
Now nine more months of data are available, and here's what we see:
Sunday, March 28, 2010
Friday, March 26, 2010
This week the Obama Administration made that scenario come true (Treasury link).
Thursday, March 25, 2010
Suppose for a moment that their claim -- unemployment time is socially productive -- is correct. Still, the claim does not rationalize unemployment insurance. Actually working is productive too, that's why people get paid for doing it. And it's widely recognized that the observed behavioral reactions occur because unemployment insurance subsidizes unemployment relative to working. Efficiency is lost whenever one productive activity is subsidized relative to another. In other words, the argument that unemployment insurance is inefficient in no way relies on an assumption that unemployment is entirely wasteful.
The marketplace involves a myriad of decisions between alternate productive activities, and the case for subsidies happens only when the market by itself would excessively favor one over the other.
So if you want to defend unemployment insurance on the grounds that unemployment is productive, you need to further argue that the market by itself results in too little unemployment, so that unemployment insurance helps ensure that we get the extra unemployment that society needs!
Such a claim would be logical, and perhaps even correct, but I predict that stimulus law advocates are unwilling to make it.
Wednesday, March 24, 2010
Macroblog reports that new firms have been doing proportionately more hiring. Comparing new and old firms is a good way to see whether its (lack of) new loans that is holding back business growth or some kind of problem with old loans. New businesses are not burdened by old loans, but they do benefit from new loans.
I have seen some commentary that the higher income people have fared better in the recession. I'm not sure that's true, but if so, that is consistent with the view that various means-tested benefits are reducing labor supply because the highest income people are unaffected by means-tested benefits.
Many auto insurance contracts have a specified value (e.g., blue book value) for auto accident awards, and that specified value may not immediately adjust to market conditions. Thus, IF it were true that used car prices fell during the recession, this would reduce the cost to an insured driver of having an accident that "totaled" his auto, which could increase the frequency of accidents that resulted in totaled cars.
Suppose, for example, your car is insured for $5k. Before the recession, a total accident would cost you $1k plus medical bills and hassle because purchasing a similar car would cost you $6k minus the $5k insurance award. During the recession, you could replace your car for $4.5k, so a total accident during the recession would give you a profit of $500, minus medical bills and hassle. So the recession made accidents cheaper (but not necessarily free).
This does not even count the fact that you might not be working during the recession, and have less of a need for your car, or might like to replace your car with one cheaper to operate. One may also want to adjust for the fact that less driving occurs during recessions: the question is whether the driving-adjusted accident rate is higher during a recession.
I have not seen any data on this, but it's something interesting for an economist to investigate.
The "Stimulus" Law of 2009 is a huge example of a public program with consequences opposite of its name.
“In Sweden, the centre-right government elected in 2006 has implemented significant income tax reductions (and payroll tax reductions), particularly for low earners while at the same time, unemployment and sick leave benefits have been cut and it has become much more difficult to be able to receive such benefits.
In Denmark, by contrast, even though it too has a formally "centre-right government" it has largely refrained from reducing taxes or government hand-outs, and it has more specifically lacked the focus of their Swedish counterparts in strengthening incentives to work by reducing unemployment benefits and taxation of low income earners.”
Denmark’s employment has fallen 5% in the past year while Sweden’s is unchanged.
CBM: I am embarrassed to report having been too busy in this recession to give much attention to the details of European recessions, but I did make one comparison of European countries that included Sweden (with a 2 percent employment gain over 7 quarters) but did not include Denmark due to lack of GDP data. The same employment data show a -0.6 percent employment loss for Denmark. However, the theme of my article was that countries that allowed employment to adjust downward may, remarkably, enjoy MORE GDP as a result.
Determining whether my article's suggestion is correct, and whether that logic has much to say about the Denmark-Sweden comparison, requires further research.
Means-tested government benefits, and the “resentment zone” that goes with them, are more prevalent than ever, thanks to the housing crash. Programs like these delay the housing market’s recovery.
Home buyers take usually out mortgages that cover only part of the value of the houses they are buying. In other words, the house is worth more than the mortgage owed. This means that in the event a borrower defaults, the lender can, in most cases, be repaid in full by foreclosing on the house and selling it to someone else.
However, housing prices have fallen dramatically since 2006. By 2009, about one in four home mortgages was “underwater” — meaning that the market value of the house had fallen below the amount owed on the mortgage (a.k.a., home equity is negative). Because of the low resale values, foreclosing on any of the homes will not yield lenders their entire principal; lenders in those cases must rely on the good behavior of the borrowers.
Meanwhile, homeowners are considering whether it is worth moving, and a reduced credit rating, in order to erase their negative home equity by walking away from their home and its mortgage.
As it has with health care, student loans, nutrition and other areas, the federal government has responded to the mortgage crisis with another layer of means-tested benefits. And these may be the largest means tests yet.
The government’s idea was to “modify” underwater mortgages in order to give homeowners an additional incentive to stay in their house. But the government will not sponsor such modifications for people with negative home equity who also earn “enough” to make their mortgage payments.
The government programs aim to modify the mortgage so that the modified mortgage payments, together with taxes and insurance, are 31 percent of the borrower’s income. Those reduced payments remain in place for five years, if not longer.
Thus, a homeowner earning $60,000 per year would have his annual housing expenses reduced to $18,600 (31 percent of his income), whereas one earning $70,000 would have expenses of $21,700. In other words, the fact that one person earns an additional $10,000 for the year obligates him to $15,500 additional housing expenses (an additional $3,100 per year for five years).
For this reason, mortgage modification has dramatically larger penalties for earning than other means-tested government programs. This not only adds to an already large pool of resentment, but is also partly responsible for the slow pace of modification by mortgage lenders and servicers, and for the dearth of income tax revenues collected by state and federal treasuries.
Tuesday, March 23, 2010
"1. While both parties included income tax, EITC, and child tax credits, we also included social security and Medicare taxes. Because we included both the employer and employee side of those taxes, we also included the employer share as income when figuring out the overall tax rates. This is the same method the JCT employed, and they explain it a bit better than I just did in footnote 1 of http://www.hudson.org/files/documents/Response_to_Mark_Prater_Effective_Marginal_Tax_Rates.pdf.
2. We included the loss of the bill's cost-sharing tax credits, which Holtz-Eakin and Brill did not. The cost-sharing tax credits are outlined on pp. 261-262 of http://democrats.senate.gov/reform/patient-protection-affordable-care-act-as-passed.pdf, but essentially, families at 150% and below of the FPL get insurance with a 90% actuarial value, families with incomes between 150 and 200% FPL get insurance with 80% actuarial value, and families above that get no special treatment.
3. We also assumed that people would purchase a 60% AV plan (called a bronze plan in the bill) rather than the subsidized 70% AV silver plan if it were cheaper for them to do so. I'm not certain if they chose to consider that or not, but it's not in their graph. (This accounts for the final spike in our charts). Depending on how expensive you predict insurance will be, families might not necessarily face this choice if they hit 400% FPL--the phase out of all subsidies--before their mandate tax rate exceeds the cost of a bronze plan. When that happens, there is an even steeper cliff at 400% FPL."
It will take me a while to digest what all this means, but in the meantime readers of this blog may want to pursue this further -- I hope the above information helps.
Monday, March 22, 2010
- Means-tested health insurance benefits
- Extra taxes for multi-national corporations (see Wall Street Journal; HT Greg Mankiw)
- mandating the employers with large payrolls provide health insurance, but that employers with small payrolls do not,
- means-tested mortgage modification (presenting millions of workers with implicit tax rates in excess of 100% (sic)),
- means-tested new home buyer credit,
- mean-tested student loan modification,
- unemployment insurance extensions,
- state income tax hikes,
- IRS means-tested enforcement of prior tax debts,
- planning for new-hiring-tax-credits,
- marginal federal tax rate hikes on the "rich"!
Both the Senate and House health bills are examples of the kinds of work disincentives (see also here) created by our own government that made employment fall so much more during this recession than it had to, and guarantees that the ratio of employment to population will for decades remain below what it was before President Obama took office. The basic problem is that people keep less of their income at the margin because not only does earning income cause them pay more state, local, and federal tax, but it also causes them to lose government health benefits.
The chart below displays estimates of marginal tax rates for various income groups, under three laws (the chart was made in January, so "current law" meant neither House nor Senate bill). This chart comes from a Cato Institute study, and was displayed on the Kaiser Foundation's web site.
I do not yet understand why the marginal tax rate hikes from the plans are significantly greater in this chart than shown in the WSJ, which cited the study above. The WSJ chart is below (has Senate plan only), and at first glance seems more credible to me (but I am checking into the differences and will post better-than-first-glance conclusions later). But both charts agree that marginal tax rates will be significantly higher.
An important part of my interpretation and forecasting of the macroeconomy was the assumption that higher marginal tax rates were here for the long term -- I have to thank lawmakers for making my forecasts come true!
Sunday, March 21, 2010
Today it was reported that now U.S. Treasuries yield about the same as some high grade corporate debt, after decades of yielding significantly less. Berkshire Hathaway debt actually yields less (but see the comment below that Bloomberg made a mistake)!
Unfortunately, Milton Friedman and economists since him do not have an empirically accurate theory of why private sector securities have such different expected returns, so it's not clear which of the many private returns would be the same as the Treasury yield in a world with the optimum quantity of government liabilities. Is the average private return? Or the returns on private securities with similar "risk" as Treasuries?
The answer depends on what is your theory of why private assets can have different returns, but arguably we have now reach the optimum quantity of government liabilities.
Thursday, March 18, 2010
Professor Robert Shimer: all benefits account for 1 - 1.5 percentage points of the current unemployment rate
Federal Reserve Board: "The several extensions of emergency unemployment benefits appear to have raised the measured unemployment rate" (Jan 2010 FOMC minutes). They cite other estimates, which could overlap with the above.
UPDATE: Econbrowser claims that raising the unemployment rate might enhance welfare -- for a more detailed discussion of that remarkable claim, see this short post.
For my own decomposition of the "offsetting" effects of UI, see this post from last summer.
I am well aware that the Meyer study and the study of Pittsburgh are "partial equilibrium" -- in theory it is conceivable that one person's work behavior is offset (or more than offset) by the work decisions of others. Indeed, I have given a great deal of attention to empirically estimating the feedback effect, and found (in this paper, this post, and the posts linked therein) that an increase in labor supply of one group is not fully offset by decreases in labor demand for other groups, but rather increases aggregate labor usage (as partial equilibrium theory would predict).
Although a bunch of economists proclaim that UI's general equilibrium effect offsets its partial equilibrium effect, they have put forward zero evidence in support.
Wednesday, March 17, 2010
Unemployment benefits provide a small amount of help to a number of people who desperately need it. But some economists have gone too far by claiming that unemployment insurance is stimulating the economy.
Unemployment insurance is jointly administered and financed by the federal and state governments, offering funds to “covered” people who lost their jobs and have as yet been unable to find and start a new job. The program has been around for decades, but this recession has created an especially large group of laid-off workers who, despite an extensive search, genuinely cannot find another job.
With no new job in sight, a large group of people are under considerable personal and financial stress. In recognition of these facts, the “stimulus” law of 2009 extended the eligibility period for unemployment benefits, and provided additional funds for the program.
Before this recession, most economists probably thought that some amount of unemployment benefits were just and compassionate, and offered a sense of security even to people who were lucky enough to retain their jobs, despite the fact that the program would raise unemployment rates and reduce both employment and economic output.
In other words, unemployment benefits shrink the economy to some degree, but shrinking the economy a bit may be a price worth paying.
Unemployment benefits were thought to reduce employment and output because, by definition, working people were ineligible for the benefits. In particular, an unemployed person who finds and starts a new job, or returns to working at his previous job, is supposed to give up his unemployment benefits. Economists had found that a large fraction of unemployed people delay going back to work solely because the unemployment insurance program was paying them for not working.
Fewer people working means a lower employment rate, and less output because unemployed people are not yet contributing to production.
The recession has seen a number of economists ignore prior findings on unemployment insurance, at least as long as this recession continues. For example, in evaluating the stimulus law economists at the nonpartisan Congressional Budget Office assumed that the law would raise gross domestic product, and took no account of the fact that the unemployment insurance and other provisions of the stimulus law give people incentives to work less.
Paul Krugman recently summarized the incentives-do-not-matter point of view in his blog (see also a leading labor economist at Harvard, quoted here):
Everyone agrees that really generous unemployment benefits, by reducing the incentive to seek jobs, can raise the [normal unemployment rate]… But in case you haven’t noticed … What’s limiting employment now is lack of demand for the things workers produce. Their incentives to seek work are, for now, irrelevant. [emphasis added]
I have not seen any evidence to support this claim that, in essence, the laws of economics are suspended as long as a recession continues. Rather, the available evidence suggests the opposite.
One feature of the unemployment insurance program is that it specifies an exhaustion date: a person stops receiving benefits when he or she starts working again, or reaches the exhaustion date (often a fixed number of weeks after being laid off), whichever comes first.
In the past, economists observed that a large fraction of unemployed people suddenly started working again within a week or so of their exhaustion date, despite having been without work for so many weeks prior: evidence that the benefits themselves were sustaining unemployment.
If the Congressional Budget Office, Professor Krugman and others were correct, this pattern would be absent during a recession, because the demand just isn’t there, and demand will not miraculously appear merely because of the arrival of the benefit exhaustion date.
A study published by two labor economists, Štepan Jurajda and Frederick J. Tannery, looked at employment histories for unemployment insurance recipients in Pittsburgh in the early 1980s. Unemployment rates got quite high in Pittsburgh in those days, reaching 16 percent at one point, and staying over 10 percent for two and a half years.
The chart below summarizes their findings for Pittsburgh.
The chart displays the fraction of persons (in Pittsburgh) receiving unemployment benefits who began working again, as a function of the number of weeks until their unemployment benefits were scheduled to be exhausted. For example, a “hazard” value of “0.04″ for week “-14″ means that, among unemployed persons with 14 weeks remaining until their benefit exhaustion date, 4 percent of them either began working a new job or returned to their previous job.
Very few people started working during the two to three weeks prior to the exhaustion of their unemployment benefits (weeks “-3″ and “-2″ in the chart). But almost 30 percent started work just a week later (19 percent started a new job, 10 percent returned to a previous job).
“Demand” may have been lacking in Pittsburgh in the early 1980s, but that did not stop unemployed people from responding to the work incentives presented to them by the unemployment insurance program (economists also looked for this pattern during a Swedish recession, and found it there too).
Unemployment insurance is only a small part of the reason why the labor market has so far failed to restore employment to pre-recession levels. But unemployment insurance is not free: It results in less employment and less output, not more. The real question is whether, and for how long, this price is worth paying to continue a just and compassionate program.
Monday, March 15, 2010
Friday, March 12, 2010
added: here's a Canadian (inflation-unadjusted?) perspective.
Thursday, March 11, 2010
Another possible link is that business asset values determine corporate tax liabilities. When assets values fall, the IRS effectively owns the asset at the margin, and the corporation has no reason to nurture it. I read an example of this in AOL's business (HT: Professor Perry): it looks like some workers at Bebo can expect to be fired solely because of the corporate tax code.
Wednesday, March 10, 2010
The elasticity has long been studied in economics; one of the seminal studies was published by Professors Topel and Rosen in 1988.
The most recent (2000-2005) housing boom was a boom in both housing prices and construction -- was it a movement (albeit, further) along the same sort of supply curve as in previous housing cycles?
Unfortunately, I have noted before that various housing price indices disagree widely about the magnitude of the housing price boom. The Case-Shiller index (adjusted for general inflation as measured by the PCE deflator), says that housing prices were 0.56 log points (about 75%) .17higher in 2005 than they were in 2000. The OFHEO index says 0.27 log points. The Census Bureau quality-adjusted index of the selling prices of new homes (the index used in the Topel-Rosen study) says 0.17 log points. The BEA price index for new residential structures says 0.15 log points.
Because the indices give such different estimates of housing price changes, they would give equally different estimates of the supply elasticity. So, in order to answer the question, it is crucial that I stick with the index used by Topel and Rosen.
Real residential investment per capita increased 0.24 log points. So that implies a supply elasticity of 1.4 for the period 2000-2005.
Topel and Rosen found significant lags in the response of investment to housing prices, whereas the boom and bust since 2000 interestingly show housing construction and housing prices to change direction in essentially the same month.
Putting that aside, Topel and Rosen find a 1.0 supply elasticity with respect to a housing price elevation that lasts one year, a 1.6 elasticity for a 4-year elevation, and a 1.7 elasticity for a 8-year elevation. So that looks pretty close to what happened in the recent boom, occurring after Topel and Rosen wrote.
[Technical note: to adjust for trends, Topel and Rosen put a linear (NOT exponential) trend in a level-of-investment regression. That is not appropriate here -- would be very sensitive to which years were included -- so I normalized investment by population]
- government purchases of mortgages, pieces of mortgages, and liabilities of funds and institutions that own mortgages or pieces of mortgages, in excess of their market value. It's the excess over market value that counts as a subsidy. For example, most of the $100+ billion the government paid to AIG was a subsidy, because what it got in return was essentially worthless (from the market's point of view). The $700 billion TARP would not be all subsidy, but the Treasury received some positive value assets in return. On the other hand, I would say that Goldman Sachs received a TARP subsidy, even though it eventually paid the money back with interest, because the government paid more than market price for its Goldman Sachs securities (e.g., in another state of the world, Goldman Sachs never pays back).
- the increase in the net worth of the owners of mortgages due to the fact that mortgage lenders are permitted to discriminate in their modifications on the basis of borrower tax returns (thereby depleting federal, state, and local Treasuries of income tax revenues).
I have seen various estimates of government monies disbursed, and committed:
- CNN's bailout tracker (updated through Nov 2009)
- The New York Times' Bailout Tab (Feb 2009)
- Bloomberg's Rescue Fund Breakdown (Dec 2008)
These amounts are $3-8 trillion. However, much of that is not a subsidy (see above), and these total entirely exclude the implicit subsidy occurring through mortgage modification.
Luke Threinen and I calculated that the U.S. housing market crashed a total of $5 - 8 trillion (difficult to know exactly, because housing price indices disagree so much on the amount of the price reduction). So it's easy to see that government paid for 25% of that loss, and conceivably paid for 50% or more.
Commenters: have you seen any studies of the amount of mortgage market subsidies?
There are two fundamental, yet universally ignored, reasons why the purchase price of a house might vary MORE than the flow of value-added.
First, it is rare that a structure by itself can provide living space to the highest value occupant -- other resources are also needed to accomplish this. Namely, real estate agents and mortgage bankers are usually needed to match an particular structure with a person or family who has a high valuation of its living space. Real estate agents and mortgage bankers share the value-added with the owner of the structure; in this view the purchase price of the house is the present value of the value-added MINUS what it costs to have real estate agents and mortgage bankers participate.
Moreover, real estate agents and mortgage bankers are much more elastically supplied in the short run than are housing structures, so an increase in the current or expected future value-added causes a less than proportionate increase in the resources spent per house on real estate agents and mortgage bankers, but (depending on the time frame over which housing structures are also elastically supplied) potentially a greater than proportionate increase in a house's purchase price.
The second source of leverage is government subsidies to the mortgage market. Through its implicit and explicit guarantees of mortgages and the banks who own them, the government injects taxpayer funds into the mortgage market when housing prices fall, but does not receive funds from the market when housing prices rise. Mortgage banking is a competitive industry, so the supply of mortgage banking shifts down when this expected subsidy is large. In particular, when high value-added (either in the present or expected in the future) temporarily raises housing prices, the expected subsidy increases (temporarily high prices = prices will fall in the future) and the supply of mortgage bankers shifts down. In this way, an increase in the demand for housing structures automatically increases the supply of resources that complement housing structures, and thereby leverage the initial demand increase.
Obviously, if you ignore these sources of leverage, you will misattribute some of the historical housing price boom to "irrationality." Perhaps more important, you will be surprised when the same fundamentals increase housing prices again.
As I’ve discussed before, national trends suggest that the sharp fall in part-time work during the last five months of 2009 can be largely attributed to last summer’s federal minimum-wage increase. A look at job changes in individual states seems to confirm this conclusion.
Two months ago, I noted how national, seasonally adjusted part-time employment increased almost 2.5 million during this recession, but then it peaked in July 2009 and headed sharply downward. I thought that it was no coincidence that the federal minimum hourly wage was raised, on July 24, 2009, from $6.55 to $7.25 — especially since the inflation-adjusted federal minimum wage had already gotten pretty high.
I estimated that part-time employment would have been about 500,000 greater in the last couple of months of the year if it hadn’t been for that last increase in the federal minimum. Many of us view 500,000 job losses as a lot because they would, for example, make a sizable dent in the jobs that the White House claimed last year that it would create with its $787 billion stimulus law.
Some of the readers of this blog, as well as the Economic Policy Institute, wondered whether my conclusion about the federal minimum could be confirmed with regional data.
After all, 19 states were not directly affected by the July 2009 federal increase because they already had their minimum wage at least at $7.25 per hour (the federal minimum prevails over the state minimum whenever the former is greater). Thus, whatever jobs were lost because of the federal increase would be lost in the other 31 states and the District of Columbia (which sets its minimum to be $1 above the federal minimum, so when the federal minimum jumps, do does D.C.’s).
Testing the state-level hypothesis with the Census Bureau’s household survey is not so easy because 500,000 job losses, significant as they are, can easily be swamped by seasonal and statistical fluctuations in the state-level household surveys.
Indeed, I noted before that even changes in the seasonally adjusted national-level household survey look a bit random from month to month. So it is no surprise that seasonally unadjusted, state-level monthly measures of part-time employment are also pretty noisy. (To my knowledge, neither the Census Bureau nor the Bureau of Labor Statistics has seasonally adjusted, state-level monthly part-time employment data.)
Another problem is that high-minimum-wage states have different labor markets than the others.
I try to alleviate this problem by comparing part-time to full-time employment, and considering just the months immediately after the federal move. (Combining a study of the federal change with studies of state minimum-wage increases could make the problem worse, because each state most likely reacts to its own economy when considering the timing of its increase; see Professor David Card’s paper on this point.)
With these cautionary notes, consider the chart below. It shows how the ratio of part-time workers to full-time workers has changed since its level in July 2009 (the last month that the federal minimum wage of $6.55 was in effect). For example, a value of “0.02” for December means that the ratio of part-time workers to full-time workers in a group of states in December was 2 percent greater than it was in July. (If you’re curious: Nationwide, there is one part-time worker for every four or five full-time workers).
The ratio increased in both the “unaffected” states (blue series) and the affected states (solid red series). As we know from the national seasonally adjusted data, the ratio did not increase nationally, so seasonality tended to increase both series. My claim is not that the federal increase would halt seasonality, but merely that it would put the ratio lower in the affected states than it would have been.
The red series for the affected states is below the blue series for the unaffected states four out of five months, which is consistent with the hypothesis that the federal increase caused part-time job losses in the affected states, but not in the others.
With all of this recession’s significant labor market problems, and the expensive federal efforts to offset them, it’s too bad that the minimum-wage law added so many people to the list of those who today cannot find jobs.
[TECHNICAL NOTE: There are a couple of ways to measure part-time in the CPS monthly files, and a couple of ways to extrapolate the sample totals to state-wide totals. Shown above are the results from the most conservative (showing the least gap between red series and blue seris) of these alternatives]
Tuesday, March 9, 2010
Here was my 2008 reasoning:
"Suppose that annual real interest rates were going to be one percentage point (100 basis points) lower for a year. Then the cost of buying a house, holding it for a year, and then selling it would be essentially one percent less. The low one-year interest rate would not affect the selling price at the end of the year because, by assumption, the reduction lasted only for a year and the next buyer will be back to normal interest rates. So the source of benefit from the low rate is that the initial buyer reduces the carrying cost for a year."
Professor DeLong was also part of Cato Issue in which I expressed the opinion above. More than a year later, he expressed the same opinion (without giving me credit, but that's beside the point):
"If you believe that the Fed kept the fed funds rate 2% below its proper Taylor-rule value for 3 years, that has a 6% impact on the price of a long-duration asset like housing. Even with a lot of positive-feedback trading built in, that’s not enough to create a big bubble."
A day later, Professor Krugman picked up on this reasoning, and nodded approvingly to Delong. All of this pointed strongly to the fact that I need to think again about my 2008 conclusion. Here's where I went wrong:
It's true that short term mortgage rates were a bit lower than normal during some of the housing boom. But the "normal" short term mortgage rate would not be the proper benchmark if the housing market really was anticipating the kinds of technical change I was writing about.
I explained how fundamentals -- the real prospects for technical change -- were temporarily pushing up housing prices. Even without subsidies, this process would efficiently raise the probability of a housing price crash, because nobody knew for sure when and how much technical progress would be realized.
But mortgages include a put option: the homeowner can trade in his house keys for the lenders' erasing his payment obligation. And the housing boom I described above (or, for that matter, any process that caused housing prices to rise and increased the probability of a crash) was increasing the value of that put option. Absent subsidies, lenders would not be giving away such a valuable put option so cheaply -- they would have charged more than normal either in the form of higher mortgage rates, higher closing costs, or lower LTVs (when multiple transactions are packaged into one "mortgage", all of these might just show up in "higher mortgage rates").
So even if short term mortgage rates had been "normal", that would have been consistent with a large subsidy, because the unusual housing price dynamics called for extra mortgage charges to reflect the enhanced value of the put option typically included with mortgages.
So anticipated government subsidies to the mortgage market magnified the housing price cycle. Prices would have gone up and come down even without them, but to a lesser magnitude. And the calculation I made in 2008, and Professors DeLong and Krugman reiterated in 2010, understates the housing price impact of those subsidies.
[Note: the elasticity of housing supply also needs to be considered (read more about that here), but that's separate from what Mulligan-DeLong-Krugman said about mortgage rates]
Monday, March 8, 2010
At the same time, I think that UI explains only a small part of the employment drop since 2007. If it were not such a great illustration of faulty economic reasoning, it would be low on my list of things to study about this recession.
- a recession that was caused solely by UI would have a larger labor force (recall that the labor force is the sum of employees plus unemployed), because some of the people who would normally leave the labor force (say, by retiring or becoming housewives) would stay in the labor force in order to get UI benefits. Our economy today has a SMALLER LABOR FORCE
- in order for a UI program to create a recession of this size, it would have to be very "well" (that is, destructively) targeted and would need a least a trillion dollars. Our UI system does not spend this much, and is not well-targeted.
- Professor Shimer has estimated that UI explains about one percentage point (maybe a bit more) of the almost 10 percentage point unemployment rate -- that leaves 9 points to be explained by something else.
- Much unemployment occurs among teenagers who are not eligible for UI.
Nor can much of the recession be blamed on the minimum wage hikes, or even the combination of minimum wage hikes and UI.
I think it is likely that employment would be millions higher if our government had abstained from the top dozen employment-reducing policies it has undertaken during this recession. But we economists have a long way to go to determine whether that policy combination destroyed 8 million jobs, or "just" 2 million (not to mention, some economists have a long way to go before they admit that government policy has destroyed any jobs).
As recently as yesterday, you could read this:
"Everyone agrees that really generous unemployment benefits, by reducing the incentive to seek jobs, can raise the NAIRU ... But in case you haven’t noticed ... What’s limiting employment now is lack of demand for the things workers produce. Their incentives to seek work are, for now, irrelevant. [emphasis added]"
And it's not just one guy. One of the world's leading labor economists was quoted as saying "Traditionally, many economists have been leery of prolonged unemployment benefits because they can reduce the incentive to seek work. But that should not be a concern now because jobs remain so scarce," and mention of incentives like these was conspicuously absent from his testimony to Congress on what's going on in the labor market. (If you think it's just two guys, look here and here).
As often as I see this claim, and how critical it is to evaluating public policy, I never see any evidence cited to support it. Meanwhile, contradictory evidence is easy to find, even during recessions. There are two types of evidence:
(a) When a group of people is suddenly presented, during a recession, with different incentives to earn, that group suddenly changes its work and earning behavior. Last week I displayed a striking chart on unemployment duration in Pittsburgh during the early 1980s recession, when unemployment was higher than it is now, and stayed that way for years. Last summer I wrote about teenagers who are much more likely seek work when school lets out for summer. Tino pointed me to a paper about the Swedish unemployment system, which became less generous during a Swedish recession (I call it a recession because unemployment had surged and surpassed 9 percent).
(b) Such groups do not merely switch places with other groups, but rather their behavior affects aggregate employment. I have shown how this worked in the summer of 2009, and for the first 12 months of this recession in the building industry.
See also "The Laws of Economics have Been Suspended!"
Saturday, March 6, 2010
By 2009, the perfect storm had formed that might even make the minimum wage noticeable from an aggregate point of view:
- The equilibrium wage likely fell due to deflation (about -2.1% July 2008 - July 2009)
- The equilibrium wage likely fell due to an increased supply of part-time workers, such as persons fired from their full-time jobs. Supply increased at least 10 percent in the quantity dimension, which pushed down the equilibrium real wage roughly 3 percent
- The nominal federal minimum wage had gotten high thanks to two recent previous hikes
- The nominal federal minimum wage was hiked again another 11 percent (from $6.55 to 7.25 per hour).
Thus, by June 2009 the federal minimum probably already exceeded the equilibrium wage for a significant number of workers, especially those part time. For example, the BLS estimates that 2.3 million part-time workers were at or below the federal minimum in 2009 (BLS does not specify which month or months, but I presume March gets a lot of weight), as compared to 1.4 million a year earlier.
So let's say that 2 million workers were affected by the 11 percent hike on July 24, 2009. With a labor demand elasticity of -3 (that's what Cobb-Douglas would predict), the textbook theory says that a half a million part-time workers would lose their jobs (or fail to be hired) due to the July 24, 2009 hike (525,177 = [1-(6.55/7.25)^3]*2,000,000).
Interestingly, the national data do suggest that about 500,000 part-time jobs were lost.
Without this perfect storm, the minimum wage would not be something of interest to macroeconomists, but "just" to observers of low-skill labor markets. Conversely, don't expect to see aggregate effects of the minimum wage unless the storm is as perfect as this one.
Friday, March 5, 2010
Thursday, March 4, 2010
In January, Diane Sawyer (this is a video podcast -- go to the 14:10 mark) explained how a couple of Reynolds 33 catamarans were thrown 50 feet in the air by a tornado/water spout/high winds in Southern California. The one shown here was called "Supply and Demand."
Fortunately, no one was on board, and no one (not even the Wicked Witch of the East) was underneath Supply and Demand when it came crashing down on the pier.
It is in remarkably good shape considering it fell 50' and had its trampoline pierced by the 12' pylon shown in the photo. But people rely on supply and demand for their safety, and this boat had enough structural damage to be considered a total loss.
Thankfully, designer Randy Reynolds says that Supply and Demand can be built again from the same specifications, and will continue to far outperform Keynesian, Marxist, Feminist, Austrian, and just about any other vessel, and do it with much greater efficiency.
Boat US and the Seaworthy Insurance Company deserve special thanks for handling this freak accident so professionally, and promptly and cheerfully paying Supply and Demand's owner pursuant to its insurance contract! Randy Reynolds and Bob Long are not only great sailors and boat builders, but also engineered a quick recovery from the storm. Thanks also to the United States Coast Guard.
If the federal government gives fungible $ to states, then the states could react either by cutting current state taxes (or increasing them less), cutting future state taxes (even states with balanced budget amendments could do this by adjusting their capital expenditures), or increasing spending.
The Obama White House estimated that 1/3 of fungible federal $ to states goes to state tax cuts, and the other 2/3 to increased spending -- a pretty good estimate IMO. The tax cuts do not necessarily have to be in the present, and do not necessarily have to reduce marginal tax rates (e.g., some of the recent federal "tax cuts" increased income-weighted marginal tax rates). The increased spending itself is very likely to raise marginal tax rates.
Thus, the effect of aid to states is probably a lot like the average federal stimulus dollar spent on other things, which raises marginal tax rates and thereby reduces GDP.
It is well known that the food stamp program acts as an implicit tax on earning, because fewer food stamp benefits are awarded to people who earn more.
This paper (Appendices B & C) calculates some marginal tax rates as a function of income, which I have translated into economy-wide average marginal tax rates from food stamps for the program BEFORE the stimulus law (the average includes zero marginal tax rates for the large number of people not eligible):
person-weighted average marginal tax rate (pre-stimulus): 4.1%
income-weighted average marginal tax rate (pre-stimulus): 1.1%
The stimulus increased the food stamp budget by 13 percent, so impact of the stimulus was to increase marginal tax rates from the food stamp program by:
0.5 percentage points (person-weighted)
0.1 percentage points (income-weighted)
The impact may seem small, but this is just a sliver of the stimulus law. If the rest of the stimulus ($766 billion in addition to the $21 billion on food stamps) were offering the same kind of incentives, the overall impact on marginal tax rates would be:
19.9 percentage points (person-weighted)
5.6 percentage points (income-weighted)
In case you think that adding 20 percentage points to marginal tax rate would not reduce employment during a recession, take a look here.
So this provision does encourage people to earn somewhere in between $6452 and $75000, rather than earning outside the range. In order to understand how that relates to "making work pay", we need to know whether it does more to pull incomes down into that range from above, as opposed to pushing them up from below.
THE NATIONAL EMPLOYMENT RATE
One question is how many people earn zero for the year, but are on the margin for working. This is difficult to know exactly but, for example, the 2009 CPS demographic supplement says that 10 percent of men aged 25-54 in 2008 earned zero that year. Only 3 percent of men and women aged 25-54 in the labor force (but not necessarily working) in March 2009 had zero earnings for 2008. So let's say that the number of persons who might consider changing their annual earnings from zero to a positive amount is equal to 5 percent of the labor force, or about 7.8 million people.
Another 20.5 million people had earnings between 0 and $6452 for calendar 2008. So we have a total of about 28 million people encourage to work, or work more.
8.4 million people earned between $75K and $95K in calendar 2008.
So clearly there are more people induced to earn more rather than less, and we can conclude that "making work pay" did push in the direction of raising the national employment rate (how much is a much more complicated question).
Note that the types of people encouraged to work more are different than the types of people encourage to work less, so it is possible that "making work pay" could reduce GDP even while it raises national employment.
The 8.4 million people earning $75K and $95K in calendar 2008 had an average income of $83K, as compared to $3K for those 20 million earning $1-6452 (who knows what the 7.8 million who earned zero would have earned if they had worked). Obviously, $400 seems like more to the latter group, and therefore would tend to motivate a greater change in hours or weeks worked for them: a rough adjustment for this compares the marginal tax rates for the two groups. The former group's marginal tax rate was 2% (ie, they lost $20 of tax credit for every $1000 they earned), and the latter group's was -6.2% (they gained 62 dollars of credit for every $1000 they earned).
So the income-weighted average marginal tax rate for the two groups was (8.4*83*.02 - 28*3*.062)/(8.4*83+28*3)= +1.1 percent. In other words, on net Making Work Pay probably REDUCED the supply of income, because its income-weighted average marginal tax rate was positive, rather than negative.
In summary, a good guess says that "Making Work Pay" may have increased national employment and hours, but it is highly unlikely that it increased GDP. The Treasury cost of the program was not only the $116 billion budgeted, but also the loss of other tax revenues due to the net disincentive to earn.
Wednesday, March 3, 2010
One instance of this would be a underwater homeowner who turns down (or fails to search for) a higher paying job in regions that are outside of the commuting distance from his house. Let's suppose that one of those regions offers a job pays well enough to justify the costs of moving.
Absent an underwater mortgage, such a homeowner would take the job, sell his house, take his equity to his new region, and purchase a house there. When underwater, but absent modification, he might still make the same choice, but would have to consider the extra cost of either (a) defaulting on or short-selling his mortgage (this would be the cost of having a lower credit score, or perhaps moral qualms with failing to pay) or (b) raising the funds to pay off his negative equity.
Absent mortgage modification, turning down the job will not make his negative equity go away. In fact, to the degree that he considers default/short-sale and his home equity is uncertain, taking that job and moving would allow him to exercise his home-value-put option while it is still in the money. In other words, persons in this situation might be too quick to move (that's the foreclosure problem).
But mortgage modification changes the story: now turning down the job and staying in the house might make the negative equity go away, because persons are eligible for federal mortgage modification only if they remain in the underwater house.
So when I say that "mortgage modification discourages people from earning too much" that does not only mean quitting one's job to "game" the mortgage modification formula, but also failing to increase one's income or take an new job when the opportunities arise.
Copyright, The New York Times Company
One of the great advances of 20th century was increased life expectancy. This advance might have bankrupted Social Security, if it were not for women in the work force.
When the Social Security program was created in the 1930s, life expectancy was 60 years, as compared to 75 years in the 1980s and 78 years today. With 1930s life expectancy, a great number of people were expected to pay Social Security taxes while they worked, but never live long enough to ever collect benefits. The early planners of the system understood that the prior contributions of the now-deceased were one way that a retiree could collect more in benefits than he paid in taxes.
Although there are still some Americans who work, pay taxes, and then die before collecting Social Security benefits, this life cycle is much less common than it used to be. Thus you would think that Social Security would have had a huge deficit years ago.
Social Security’s architects did not anticipate another major advance in the 20th century: women’s progress in market work, and how that progress would fill Social Security’s coffers.
Once they turn 62 years old, married people, widows and widowers are eligible for Social Security, regardless of whether they ever worked for pay during their lifetimes. When Social Security was in the planning stages, it was expected that most working-age couples would have only one person working (and thereby paying the payroll tax) — typically the husband. Yet when these couples retired, Social Security would pay benefit checks to both of them as long as they lived (see the chapter in this book written by Professors Jeffrey Miron and David Weil, for more on the economic and demographic projections originally used to design Social Security).
Thus, a payroll tax rate was planned that could bring in enough revenue from working men to pay both the retired men and the retired women.
History did not quite turn out that way. In fact, millions of married women worked for pay and paid the payroll taxes as they did. This was largely profit for the Social Security system, because the system would have paid those women benefits regardless.
The revenues of governments in the future will depend just as much on how women spend their time. Governments can expect more revenue if women continue significantly with their labor market progress, and less revenue if some of women’s payroll gains are reversed in the years ahead.
Tuesday, March 2, 2010
The law gives the most per capita to states with the least unemployment. Read more here.
added: The White House appears to read economix, and yesterday offered a rebuttal of Professor Glaeser's critique.
As you know, I write weekly at economix, so we can understand the White House's complete silence regarding my critiques to mean that they do not have much to say:
Obviously, they agree with "Who's Afraid of the Big Bad Deficit?" and "More Government Debt, Please."
In case you are wondering where else the CBO went wrong, look here. In short, they completely ignore incentives even while a long economics literature has shown that they have real effects, even in recessions.
Monday, March 1, 2010
However, it does not address the single biggest criticism I have made on this blog, on TV, and elsewhere: that much of the stimulus spending (and many of the so-called tax cuts) goes only to persons and businesses in financial hardship, and thereby serves as a tax on success. In other words, stimulus spending is an implicit income tax, and thereby reduces national income rather than increasing it.
Implicit taxes have been familiar to economists for decades, and are widely acknowledged to be a big part of the economics of pensions, unemployment insurance, welfare, and other programs: the CBO and so many other economists have no excuse for ignoring them.
The CBO claims to be critical of general equilibrium models (although it approvingly cites a bunch New Keynesian general equilibrium models), but the problem with implicit taxes has nothing to do with general equilibrium reasoning.
The CBO also offers some critiques of "rational models", although the problem with implicit taxes has nothing to do with rationality, either.