This is Blog Author's Typepad Profile.
Join Typepad and start following Blog Author's activity
Join Now!
Already a member? Sign In
Blog Author
Recent Activity
Ric: We formatted the graphs in such a way to best compare the size of the discontinuity rather than the overall scale of the graphs, but the graphs of the discontinuities for ages 15-34 and 35-59 have slightly different scales. In fact, the number of cases per capita for 35-59 year olds is somewhat higher than that for 15-34 year olds on average and at the border. Furthermore, out of the 400 German counties included in the analysis, in 301 counties the number of cases per capita for the older group exceeds that for the younger group. While the COVID-19 epidemic in Germany indeed was concentrated among the young to a greater extent than in multiple other countries, Germany as a whole does not appear to be an exception to the medical community consensus that you cited.
Dino: We focused on the case of Germany because there are smaller differences across the border as both sides have the same central government. Looking at Spain vs. Portugal (or Northern Ireland vs. Ireland, where BCG-vaccinating Ireland has more cases per capita than its non-vaccinating neighbor) would introduce more confounding factors into the analysis.
Daniel: Thank you for your question. When you have multiple jobs and lose one of them (or experience reduced hours at either or both), you can file a partial loss claim. Whether or not you qualify for unemployment benefits still depends on your state’s eligibility criteria, and so there isn’t much difference there in most states. However, the weekly amount you will receive conditional on qualifying depends on the amount you are still earning. Again, the exact amount you will get varies across states. How do such job losses factor in the official statistics? The payroll survey reports the net change in number of jobs. Let’s take a person with two jobs and suppose that a person loses both of them and does not find another work. This would count as two jobs lost on net in the payroll (establishment) survey data. The employment and unemployment rate statistics in the household survey, however, are based on the number of people, not of jobs. Therefore, in the example above, the person would count as one unemployed.
Frank: Thank you for the question. Continuing claims refer to unemployed workers that are currently collecting unemployment insurance benefits. Therefore, while initial claims is a proxy for people that lose their job, continuing claims is aimed at capturing the evolution of the stock of unemployed workers. The relationship between continuing claims (the insured unemployment rate) and the “true” unemployment rate is also a bit murky in general. First of all, continuing claims does not capture people that are unemployed but do not qualify for benefits, although this discrepancy would likely be much smaller this time, since the CARES Act relaxed eligibility criteria. Second, there are outflows from continuing claims that are not related to going back to work but are simply due to benefit exhaustion. However, this is not relevant now, as it will be some time before recently laid-off workers start seeing their benefits expire. On the other hand, there is one way in which changes in continuing claims are a more accurate indicator of changes in the labor market than initial jobless claims: that is that they reflect net, rather than gross changes—i.e., when someone in that group finds a job, that is reflected in the continuing claims numbers (they go down), but not in the initial claims numbers. Consistent with this idea, the weekly change in continuing claims is typically well below the level of weekly initial claims. We hope this is helpful.
Fred: Thank you for your comments. The effect of this pandemic on labor force participation—both in terms of magnitude and persistence—is clearly an open question. As you suggest, it may be a negative effect (as in #1 and #3 above) or a positive effect (as in #2 above). At any rate, if this does end up discouraging some people out of the workforce, that might adversely affect the level of employment, but it would probably have a downward effect on the unemployment rate (and an upward effect on wages). Only time will tell. --------------------- Phil: Thanks for your comment. Your stool comparison is an interesting one. On the other hand, there may well be **more** impetus to rebuild this time, because migrating to an unaffected place (like Houston after Katrina) is simply not an option. To illustrate this, there was a spike of about 150,000 in Texas’ population in 2006—75,000 in Harris County (Houston) alone—which appears to reflect outmigration from Louisiana. This is unlikely to happen this time, at least to anywhere near the same extent. --------------------- Buddy: Thanks for your comment. This is an excellent point. It is true that New Orleans was hit much harder than the rest of the state, especially in terms of flooding, which was the major problem. However, jobless claims—the only really solid economic series with recent information on COVID19—is only available at the state level. It is true that the New Orleans metro saw a steeper percentage job loss, and took even longer to come back. However, there was a good deal of variation across Louisiana, and there is likely to be a good deal of variation during this pandemic across the U.S. At any rate, while it is unclear if Louisiana or New Orleans is the better comparison, that should become clearer in the weeks ahead as we get more regional economic data. Our main point, though, is that comparisons to these types of natural disasters are more informative than comparisons to the 2008-09 recession.
In reply to YD: To scale the WEI to GDP growth, we compute the average WEI over a quarter and report the predicted values from a regression of GDP growth on the “quarterly WEI” and a constant.
In reply to Woodward: Regular updates are posted at We anticipate the WEI being available from FRED in the near future.
In reply to Leahey: We are aware of the Homebase labor market data. We are continuing to evaluate potential series and may refine the index in the coming weeks.
Thanks for your interest. We anticipate making this data public shortly.
Thanks for your question. Existing work both within the Federal Reserve System and outside shows results consistent with this finding. The Atlanta FED wage tracker has been showing similar results for a while. Here is a link to one such article: The New York Times also ran a similar story recently relating this to minimum wage hikes. Here: The authors look at this in a slightly different way, specifically they link same people over a 1-year window and study their wage growth. They find that wage growth at the bottom has always been higher, but this is also to be expected because low wages also reflects transitory factors (as discussed in more detail in the post). The earnings measure includes base salary or wage, and tips, bonuses, and overtime compensation. But it does not capture fringe benefits or stock options.
Ted: thank you very much for your questions. On your first question, in the associated working paper we controlled for the level of subprime lending in 2002, and the results we report in this post still hold. Also, the initial level of subprime lending is not correlated with subsequent home price appreciation. Regarding your second question, in the paper we show that the results hold at a zip-code level. We also note that most of the variation in home price appreciation in our sample period was at the county level and that the variation in home price appreciation across zip codes within counties was relatively small. Finally, thank you for your note on the footnote of the subprime map! We have updated it.
Jared, thanks for your questions. On your first question, we did find some evidence that risky loans were decoupled from subprime borrowers. Relatedly, we also found some evidence that loans to subprime zip codes within boom counties were decoupled from subprime borrowers; that is, prime borrowers were increasingly buying in subprime zip codes within boom areas, potentially pricing out subprime borrowers. Regarding your second question, we have not looked at higher-priced loans specifically, but that could be something for future research. Finally, the growth of foreclosures was more concentrated in home price boom counties rather than in subprime boom counties.
Thank you for your comments. Yes, the evidence suggests that the new tick size is often binding, albeit to a lesser extent than the old tick size. That said, there is evidence from other markets that the tick size can be too small (see our underlying New York Fed Staff report for more details), with a smaller tick size reducing the incentive for liquidity provision and possibly harming liquidity. It’s therefore not necessarily the case that smaller is always better. Moreover, determining the optimal tick size is challenging because liquidity has different aspects and is always changing.
Scott: We argue that U.S. firms and consumers are paying the higher tariffs because Chinese firms have not lowered their prices to offset the tariffs. We note that it is difficult to break down how much of the tariff bill is ultimately being paid by U.S. firms and how much by U.S. consumers. It is quite possible that U.S. consumers are paying a small share, with the cost of tariffs largely reflected in lower profit margins for U.S. firms.
Sara: Indeed, degree completion and major choice will all have an impact on an individual’s income (and thus ability to repay their student loans). Unfortunately, these factors are not observed in our data set.
Ira S: Race can affect outcomes in many ways, including ones you mention. We are simply reporting the relationship between race and outcomes.
Curious George: There actually isn’t much variation by *lender* in the student loan space, since more than 90 percent of the outstanding student loan balance is federal loans (thus the lender, for virtually all, is the federal government). There is some variation in which firms service the loans. Our data don’t reveal lenders or servicers.
In reply to Kenneth Dreifus: Thanks for your comment. The main reason we selected those two sectors—retail trade and leisure & hospitality—was that they tend to be the lowest-wage industries and thus most susceptible to an increasing minimum wage. To illustrate this, at the national level, the average hourly wage of a manufacturing worker was $24.80 in 2014, versus $17 in retail trade and just under $14 in leisure & hospitality. Clearly, there are many manufacturing workers making less than that, but we wanted to gauge the effect on those industries with the largest exposure. Still, it is true that, as you suggest, manufacturers would be much more flexible about location than retailers or restaurants. And it may well be the case that the manufacturing sector has seen some employment effect from the minimum wage hikes. This is an area of research that we intend to pursue going forward. However, bear in mind that the fact that people may not cross state lines for a cup of coffee does not, by any means, rule out the potential for adverse employment effects—restaurants, hotels, and retailers can invest in automation, cut back on more labor-intensive services, and implement other such labor-reducing policies … or they may simply go out of business.
Jeffrey: Many thanks for your comments on this post. In terms of the spatial unit of this analysis, we focus on metropolitan areas as they are a good approximation of local labor markets. So, yes, when we refer to Fairfield, CT, in this post it is the Fairfield County metro area, not the town.
In reply to Or-el: Thanks for your interest in our post and thanks for your question. In response, we did not make any inflation adjustment to earnings nor do we believe this is a concern for a few reasons: 1) There are no broad measures of inflation at available at such a localized level. 2) Even if such measures were available, CPI inflation typically does not vary much geographically—even across widely different metro areas in the United States, let alone adjacent counties. 3) This analysis is based on where people work, and it is likely that at least some people who work on the New York side of the border live in Pennsylvania and vice versa. So even if the inflation rates were slightly different, any effects would be muted.
Jeffery: Thank you for your interest and your thoughtful questions. In this analysis, we focused only on employment and earnings, not the number of establishments. But I think it is almost certain that the USVI did indeed lose a larger proportion of such establishments, after the storms, than Puerto Rico. If so, the same question remains: why? Insofar as disruptions in the USVI (positively) affected Puerto Rico, it’s important to remember that PR’s economy and population are about 30x as large as the U.S. Virgin Islands’, so it’s unlikely that anything going on in the USVI would have any significant direct effect on Puerto Rico. As for the notion that some of the high-paying construction jobs may be going to workers imported from the mainland, that is indeed a possibility in both USVI and Puerto Rico—and that would cause *both* the post-hurricane earnings and employment levels to be somewhat overstated. However, to be included in the islands’ job and income counts, a worker has to be on the payroll of a local employer; contract workers from the mainland are not included. And finally the policy ratcheting up the minimum wage in the U.S. Virgin Islands was implemented long before Irma and Maria, so the storms would not have been a factor in that decision.
Peter: One key advantage of our transition rates is they measure the flow into delinquency -- thus the non-dischargeability is not an issue. It is true that the transition rate on student loans has not improved with the recovery, which is surprising especially with the multitude of repayment plans available to borrowers now. We are currently working on analyses on this topic (stay tuned!).
Stefan: Thank you for your comment and for reading our blog. The effective fed funds rate can print below interest on reserves (IOR) because government-sponsored enterprises (GSEs) are eligible to lend funds in the federal funds market, but are not eligible to earn IOR.
Wojciech: Thank you for your questions and for your interest in our work. Interest on reserves (IOR) has not served as a hard minimum rate at which all institutions are willing to lend funds because some institutions are eligible to lend funds in the federal funds market but are not eligible to earn IOR, such as the government-sponsored enterprises (GSEs). Clicking on “fed funds lending” in our interactive map shows that lending in this market could be done by two types of entities: banks and GSEs such as Federal Home Loan Banks (see for details). The rates at which these entities are willing to lend are different because banks are eligible to earn IOR while GSEs are not. In an environment with abundant reserves, banks might have an incentive to borrow at rates below IOR from an entity that does not earn IOR to then hold the funds in their reserve account and earn IOR on those funds. This would reflect arbitrage activity rather than a real funding need and could lead the effective fed funds rate (EFFR) to print below IOR (see for additional detail). In addition to IOR eligibility, other factors including the impact on their balance sheet influence the rates some banks are willing to pay for fed funds, which may limit on the rate a bank is willing to pay.
Many thanks for your comment, AG. You raise a good point on the potential influence of intraday liquidity buffers. While we didn’t explore this in our most recent analysis, this could represent an interesting question for future analysis. Relatedly, you might like to see the Liberty Street Economics post “Stressed Outflows and the Supply of Central Bank Reserves” (2/20/19), which analyzes one-day stress horizons for LCR cash outflows (it does not cover RLEN/MOL, however).