The Atlanta Fed's macroblog provides commentary and analysis on economic topics including monetary policy, macroeconomic developments, inflation, labor economics, and financial issues.
- BLS Handbook of Methods
- Bureau of Economic Analysis
- Bureau of Labor Statistics
- Congressional Budget Office
- Economic Data - FRED® II, St. Louis Fed
- Office of Management and Budget
- Statistics: Releases and Historical Data, Board of Governors
- U.S. Census Bureau Economic Programs
- White House Economic Statistics Briefing Room
November 06, 2017
Building a Better Model: Introducing Changes to GDPNow
Among the frequently asked questions on GDPNow's web page is this one:
Is any judgment used to adjust the forecasts? Our answer:
No. Once the GDPNow model begins forecasting GDP growth for a particular quarter, the code will not be adjusted until after the "advance" estimate. If we improve the model over time, we will roll out changes right after the "advance" estimate so that forecasts for the subsequent quarter use a fixed methodology for their entire evolution.
This macroblog post enumerates a number of minor changes to GDPNow that were implemented on October 30, when it began forecasting fourth-quarter real gross domestic product (GDP) growth. Here is a summary of the changes, intended to improve the accuracy of the GDP subcomponent forecasts:
- Services personal consumption expenditures (PCE). Use industrial production of electric and gas utilities to nowcast real PCE on electricity and natural gas. Use international trade data on travel services to forecast revisions to related PCE travel data.
- Real business equipment investment. Use/forecast data from the advance U.S. Census Bureau reports on durable manufacturing and international trade in goods that, previously, hadn't been utilized until the full reports on manufacturing and/or international trade .
- Real nonresidential structures investment. Replace a discontinued seasonally adjusted producer price index for "Steel mill products: Steel pipe and tube" with a nonseasonally adjusted version. The index is used to construct a price deflator for private monthly nonresidential construction spending.
- Real residential investment. Use employment data for production and nonsupervisory employees of residential remodelers to help forecast real investment in residential improvements.
- Real change in private inventories. Use published monthly inventory levels in the U.S. Bureau of Economic Analysis's underlying detail tables 1BU and 1BUC after the third-release GDP estimate from the prior quarter to estimate inventory levels for a number of industries in the first month of the quarter forecasted by GDPNow.
- Federal, state, and local government spending. Forecast investment in intellectual property products for these subcomponents using autoregression models.
The first three columns of the following table decompose the official estimate of the third-quarter real GDP growth rate, and forecasts of the growth rate from the discontinued and modified versions of GDPNow, into percentage point contributions from the subcomponents of GDP.
As the table shows, the methodological changes did not have much of an impact on the final third-quarter subcomponent forecasts—apart from inventory investment, where the modifications lowered the contribution to growth from 0.80 percentage points to 0.60 percentage points—or on their accuracy. Nevertheless, the topline GDP forecast of the modified model (2.3 percent) was less accurate than the previous version (2.5 percent). In the discontinued version of GDPNow, an overestimate of the inventory investment contribution to growth partly canceled out underestimated contributions from each of net exports, government spending, and nonresidential fixed investment.
In the modified version, the inventory contribution was also underestimated and did not cancel out these other errors. The last two columns of the table show that all of the subcomponent errors of the modified model were at least as small as their historical average for the discontinued version. However, the topline GDP forecast was less accurate than average because of less cancellation of the subcomponent errors than usual. We hope that the cancellation of subcomponent errors in the modified model will be more similar to the historical average in the discontinued version in the future.
Although the methodological changes could have more of an impact than the table suggests, we do not expect them to have a substantial impact in general. For example, on October 30, the discontinued version of GDPNow projected 3.0 percent GDP growth in the fourth quarter, which was little different from the modified model forecast of 2.9 percent growth. We provide a more detailed explanation of the changes to GDPNow here . Going forward, this same document will document any further changes to the model and when we made them.
October 19, 2017
How Ill a Wind? Hurricanes' Impacts on Employment and Earnings
According to the Current Employment Statistics payroll survey, seasonally adjusted nonfarm payroll employment declined 33,000 in September. This decline was the first drop in employment since 2010 and followed a 169,000 gain in August. At the same time, seasonally adjusted average hourly earnings in the private sector increased 2.9 percent year over year in September. This increase in average wages was the largest since the end of the Great Recession in 2009. However, it seems likely that the decline in employment contributed to the rise in average hourly earnings. Why would a decline in employment contribute to an increase in average hourly earnings? We're glad you asked!
As noted by the U.S. Bureau of Labor Statistics, Hurricanes Harvey and Irma reduced employment in the payroll survey, whose reference period is the pay period that includes the 12th of the month. Hurricane Harvey first made landfall in east Texas on August 25 and again in Louisiana on August 30, and Hurricane Irma made landfall in south Florida on September 10. The storms forced large-scale evacuations and severely damaged many homes and businesses. For workers who are not paid when they miss work, being unable to work during the surveyed pay period means they are not counted in September payrolls.
To measure the size of Harvey and Irma's effect on payroll employment, we first looked at data from the Current Population Survey (CPS). We found that the bad weather forced about 1.5 million nonfarm workers who had a job during the September reference week to miss work. Of those, about 1.2 million were wage and salary earners, and about 760,000 of those were unpaid during their absence from work.
Our analysis indicates that September saw a shortfall in seasonally adjusted payroll employment between 200,000 and 300,000 jobs, suggesting that workers returning to work could result in a large rebound in payroll employment. (Not to get too far into the weeds, but our analysis involved regressing payroll employment growth on its lagged values as well as current and lagged seasonally adjusted changes in shares of workers who were not at work because of bad weather.)
What about average hourly earnings? Changes in average hourly earnings over time reflect both the effect of people getting pay raises and changes in who is working this month versus last month or last year. This latter effect can be large during recessions, when workers in lower-wage jobs are disproportionately more likely to be laid off. The absence of these workers from payrolls increases the average wage among the remaining employed workers, even if those remaining workers are not getting much of a pay increase (see this macroblog post for more discussion).
The September payroll survey depicted a particularly large decline in employment in the leisure and hospitality sector, which is significant because average hourly earnings in that sector are typically about 40 percent lower than overall average hourly earnings. In addition, from the CPS we see that the usual hourly earnings of workers not at work because of bad weather is much lower than for other workers. These data suggest that temporary absences from work because of bad weather likely put upward pressure on average hourly earnings, and some of that upward pressure could reverse itself as these workers return to their jobs. If the pace of average hourly earnings doesn't relax, however, then that would suggest more workers getting larger pay raises due to a tightening labor market.
September 08, 2017
When Health Insurance and Its Financial Cushion Disappear
Personal health care costs can skyrocket with a new diagnosis or accident, often leading to catastrophic financial costs for people. Health insurance plays an important role in protecting individuals from unexpected large financial shocks as a result of adverse health events. Just as homeowner's insurance helps protect you from financial devastation if your house burns down, health insurance helps protects you from burning through your savings because of a heart attack. This 2008 report from the Commonwealth Fund shows that the uninsured are far more likely to have to use their savings and reduce other types of spending to pay medical bills.
Much research has been done on the impact of health insurance on financial and health outcomes. (This paper , for example, summarizes the history and impact of Medicaid.) However, most of the studies look at the case of individuals who are gaining health insurance. In a recent Atlanta Fed working paper and the related podcast episode , we measure the impact of losing public health insurance on measures of financial well-being such as credit scores, delinquent debt eligible to be sent to debt collectors, and bankruptcies. We performed these measurements by studying the case of Tennessee's Medicaid program, known as TennCare, in the mid-2000s. At that time, a large statewide Medicaid expansion that began in the 1990s ran into financial difficulties and was scaled back. As the following chart shows, some 170,000 individuals were removed from TennCare rolls between 2005 and 2006.
Our analysis of this episode, using data from the New York Fed's Consumer Credit Panel/Equifax, revealed some striking findings. Individuals who lost health insurance experienced lower credit scores, more debt eligible to be sent to collections, and a higher incidence of bankruptcy. Those who were already financially vulnerable suffered the worst. In particular, individuals who already had poor credit, as measured by Fannie Mae's lowest creditworthiness categories , and then lost Medicaid see their credit scores fall by close to 40 points on average and are almost 17 percent more likely to have their debt sent to collection agencies. Our analysis also finds that gaining or losing health insurance is not symmetric in its impact—losing insurance has larger negative financial effects than the positive financial impacts of gaining insurance.
Our results provide evidence that losing Medicaid coverage not only removes inexpensive access to health care but also eliminates an important layer of financial protection. A cost-benefit analysis of proposed cuts to Medicaid coverage (see here, here, and here for a discussion of recent legislative efforts in the U.S. Congress) would need to consider the negative financial consequences for individuals of the type that we have identified.
September 07, 2017
What Is the "Right" Policy Rate?
What is the right monetary policy rate? The Cleveland Fed, via Michael Derby in the Wall Street Journal, provides one answer—or rather, one set of answers:
The various flavors of monetary policy rules now out there offer formulas that suggest an ideal setting for policy based on economic variables. The best known of these is the Taylor Rule, named for Stanford University's John Taylor, its author. Economists have produced numerous variations on the Taylor Rule that don't always offer a similar story...
There is no agreement in the research literature on a single "best" rule, and different rules can sometimes generate very different values for the federal funds rate, both for the present and for the future, the Cleveland Fed said. Looking across multiple economic forecasts helps to capture some of the uncertainty surrounding the economic outlook and, by extension, monetary policy prospects.
Agreed, and this is the philosophy behind both the Cleveland Fed's calculations based on Seven Simple Monetary Policy Rules and our own Taylor Rule Utility. These two tools complement one another nicely: Cleveland's version emphasizes forecasts for the federal funds rate over different rules and Atlanta's utility focuses on the current setting of the rate over a (different, but overlapping) set of rules for a variety of the key variables that appear in the Taylor Rule (namely, the resource gap, the inflation gap, and the "neutral" policy rate). We update the Taylor Rule Utility twice a month after Consumer Price Index and Personal Income and Outlays reports and use a variety of survey- and model-based nowcasts to fill in yet-to-be released source data for the latest quarter.
We're introducing an enhancement to our Taylor Rule utility page, a "heatmap" that allows the construction of a color-coded view of Taylor Rule prescriptions (relative to a selected benchmark) for five different measures of the resource gap and five different measures of the neutral policy rate. We find the heatmap is a useful way to quickly compare the actual fed funds rate with current prescriptions for the rate from a relatively large number of rules.
In constructing the heatmap, users have options on measuring the inflation gap and setting the value of the "smoothing parameter" in the policy rule, as well establishing the weight placed on the resource gap and the benchmark against which the policy rule is compared. (The inflation gap is the difference between actual inflation and the Federal Open Market Committee's 2 percent longer-term objective. The smoothing parameter is the degree to which the rule is inertial, meaning that it puts weight on maintaining the fed funds rate at its previous value.)
For example, assume we (a) measure inflation using the four-quarter change in the core personal consumption expenditures price index; (b) put a weight of 1 on the resource gap (that is, specify the rule so that a percentage point change in the resource gap implies a 1 percentage point change in the rule's prescribed rate); and (c) specify that the policy rule is not inertial (that is, it places no weight on last period's policy rate). Below is the heatmap corresponding to this policy rule specification, comparing the rules prescription to the current midpoint of the fed funds rate target range:
We should note that all of the terms in the heatmap are described in detail in the "Overview of Data" and "Detailed Description of Data" tabs on the Taylor Rule Utility page. In short, U-3 (the standard unemployment rate) and U-6 are measures of labor underutilization defined here. We introduced ZPOP, the utilization-to-population ratio, in this macroblog post. "Emp-Pop" is the employment-population ratio. The natural (real) interest rate is denoted by r*. The abbreviations for the last three row labels denote estimates of r* from Kathryn Holston, Thomas Laubach, and John C. Williams, Thomas Laubach and John C. Williams, and Thomas Lubik and Christian Matthes.
The color coding (described on the webpage) should be somewhat intuitive. Shades of red mean the midpoint of the current policy rate range is at least 25 basis points above the rule prescription, shades of green mean that the midpoint is more than 25 basis points below the prescription, and shades of white mean the midpoint is within 25 basis points of the rule.
The heatmap above has "variations on the Taylor Rule that don't always offer a similar story" because the colors range from a shade of red to shades of green. But certain themes do emerge. If, for example, you believe that the neutral real rate of interest is quite low (the Laubach-Williams and Lubik-Mathes estimates in the bottom two rows are −0.22 and −0.06) your belief about the magnitude of the resource gap would be critical to determining whether this particular rule suggests that the policy rate is already too high, has a bit more room to increase, or is just about right. On the other hand, if you are an adherent of the original Taylor Rule and its assumption that a long-run neutral rate of 2 percent (the top row of the chart) is the right way to think about policy, there isn't much ambiguity to the conclusion that the current rate is well below what the rule indicates.
"[D]ifferent rules can sometimes generate very different values for the federal funds rate, both for the present and for the future." Indeed.
August 30, 2017
Is Poor Health Hindering Economic Growth?
It is well known that poor health is bad for an individual's income, partially because it can lower the propensity to participate in the labor market. In fact, 5.4 percent of prime-age individuals (those 25–54 years old) reported being too sick or disabled to work in the second quarter of 2017. This is the most commonly cited reason prime-age men do not want a job, and for prime-age women, it is the second most often cited reason behind family responsibilities (see the chart). (Throughout this article, I use the measure "not wanting a job because of poor health or disability" as a proxy for serious health problems.)
In addition to being prevalent, the share of the prime-age population citing poor health or disability as the main reason for not wanting a job has increased significantly during the past two decades and tends to be higher among those with less education (see the chart).
Yet by some standards, the health of Americans is improving. For example, compared to two decades ago the average American is living two years longer, and the likelihood of dying from cancer or cardiovascular disease has fallen. These specific outcomes, however, may have more to do with improvements in the treatment of chronic disease (and the resulting reduction in mortality rates) than improvements in the incidence of health problems.
Another puzzle—which is perhaps also a clue—is the considerable variation across states in the rates of being too sick or disabled to work. For example, people living in Mississippi, Alabama, Kentucky, or West Virginia in 2016 were more than three times likelier to indicate being too sick or disabled to work than residents of Utah, North Dakota, Iowa, or Minnesota (see the maps below).
This cross-state variation is useful because it allows state-by-state comparisons of the prevalence of specific health problems. Among a list of more than 30 health indicators, the two factors that most correlate with the share of a state's population too sick or disabled to work were high blood pressure (a correlation of 0.86) and diabetes (a correlation of 0.83). Both of these conditions are associated with risk factors such as family history, race, inactivity, poor diet, and obesity. Both of these health issues have increased significantly on a national basis in recent years.
So how might poor health hinder economic growth? Health factors account for a significant part of the decline in labor force participation since at least the late 1990s. After controlling for demographic changes, the share of people too sick or disabled to work is about 1.6 percentage points higher today than it was two decades ago (see the interactive charts on our website). Other things equal, if this trend reversed itself during the next year, it could increase the workforce by up to 4 million people, and add around 2.6 percentage points to gross domestic product (calculated using our Labor Market Sliders).
Of course, such a sudden and large reversal in health is highly unlikely. Nonetheless, significant improvements to the health of the working-age population would help lessen the drag on growth of the labor supply coming from an aging population. Public policy efforts centered on both prevention and treatment of work-impeding health conditions could play an important role in bolstering the nation's workforce.
July 31, 2017
Behind the Increase in Prime-Age Labor Force Participation
Prime-age labor force participation has been on a tear recently. Over the last eight quarters, it is up by about 65 basis points (bps) and more than 40 bps in just the last year. When combined with declines in the rate of unemployment, this increase has helped lift the employment-to-population (EPOP) ratio for this key population group by around 120 bps during the last two years.
Placed in the context of an almost 260 bp decline in the prime-age EPOP ratio between 2007 and 2015, this development is significant. Although the unemployment rate is close to what most economists consider full employment, rising labor force participation can indicate that the labor market might still have some room to run before the employment gap is fully closed. (The Congressional Budget Office offers some analysis consistent with this idea.)
So what's behind the increase in prime-age (defined as people between 25 and 54) participation in the last year? Changes in the labor force participation rate (LFPR) either can be the result of changes in the mix of demographic groups in the population with different average rates of participation (for example, across education and race/ethnicity), or they can result from changes in average participation rates within demographic groups. It turns out that most of the increase in the prime-age LFPR has been because of increased LFPR within demographic groups—in particular, prime-age women and especially women without a college degree. Prime-age men have not contributed much to the rise in participation beyond the increased participation associated with a more educated population.
The following chart shows the contribution to the change in the prime-age LFPR over the last year as a result of changes in the relative mix of age-education-race groups (the blue bars) and changes in participation rates within age-education-race groups (the orange bars). It shows the contribution from both sexes combined and from prime-age women and men separately.
Note that the we computed the contributions using six five-year age groups, three education groups (less than high school, high school but no college degree, and college degree), three race/ethnicity groups (Hispanic, non-Hispanic black, and non-Hispanic white/other), and two sexes.
Of the total increase in the prime-age LFPR, most of that was the result of changes in labor force participation behavior within female demographic groups. In fact, changes in LFPR behavior from prime-age men served as a drag on the overall prime-age LFPR. The modestly positive demographic effect on the LFPR for both men and women reflects the higher LFPR for those with a college degree and the relative increase in the share of both prime-age men and women with a college degree.
This development stands in contrast to the drivers of the change in the prime-age LFPR between 2015 and 2016. Of the 24 bp increase in prime-age LFPR between the second quarters of 2015 and 2016, changes in the demographic composition of the population (primarily increased education levels) accounted for all of it rather than changes in average participation rates within demographic groups.
The next chart shows the contribution to the change in the prime-age LFPR between 2016 and 2017 due to changes in the LFPR behavior of women for specific education-race groups.
As the chart shows, the bulk of the demographically adjusted contribution from female labor force participation came from women without a college degree, and the largest contribution across female education-race groups was from Hispanics without a college degree. The increase in labor force participation among women with less education is consistent with evidence of recent improvement in the wage gains for relatively low-wage earners.
Although this simple decomposition doesn't explain why nondegreed women are increasingly finding the labor force to be an attractive option, we can infer some clues by looking at changes in the reasons people give for not participating. In particular, the largest contribution from changes in behavior among prime-age women over the last year came from a decrease in the propensity to be out of the labor force because of poor health or being in the shadow labor force (wanting a job but not looking).
Recently, former Minneapolis Fed President Narayana Kocherlakota has argued that macroeconomists should take more seriously the differences in behavior across demographic groups. The Atlanta Fed's Labor Force Dynamics web page contains more information on the behavioral trends in the reasons people give for not participating in the labor force across demographic groups, and the page was just updated to include data for the second quarter of 2017. Check it out, and we'll keep reporting here on the relative contributions to the labor force of behavioral versus demographic changes—and whether the winning streak for prime-age labor force participation continues.
July 12, 2017
An Update on Labor Force Participation
With the unemployment rate essentially back to prerecession levels, economists have been paying increased attention to the labor force participation rate (LFPR). Many economists, including those at the Congressional Budget Office , believe untapped resources remain on the sidelines of the labor market.
What exactly does "on the sidelines" entail? Discouraged workers are only a small part of the story. To help unravel the rest of the mystery behind the elevated share of people not participating, we at the Atlanta Fed use the microdata from the Current Population Survey to code the activities of persons not in the labor force. We then calculate how changes in each activity contribute to the total change in the LFPR.
The chart below depicts the drivers of the change in the LFPR from the first quarter of 2016 to the first quarter of 2017. (The interactive tool on our website allows you to make comparisons across gender, age group, and time.) The LFPR rose just slightly (about 0.06 percentage points). However, that small change was the net result of much larger countervailing forces. Other things equal, demographic changes during the year would have lowered the LFPR by around 0.14 percentage points. The aging of the population put significant downward pressure on the LFPR (pushing it down 0.24 percentage points), but a more educated workforce helped push up the LFPR (0.10 percentage points). If the age and education mix of the population had not changed, the LFP rate would have risen by about 0.19 percentage points (see the chart).
The following chart further breaks down the behavioral and cyclical components at work. After controlling for shifts in the demographic mix of the population during the year, the largest contributing factor was a decline in the rate of nonparticipation because of family responsibilities.
This is a particularly important explanation for prime-age women (defined as women between 25 and 54 years of age). A smaller share of prime-age women who say they are busy with home and/or family responsibilities accounts for about half of the 0.62 percentage point increase in LFPR that occurred between the first quarter of 2016 and the first quarter of 2017 (see the chart).
To examine factors affecting prime-age men's participation or to learn more about the cyclical and structural factors behind each reason, visit our website.
July 11, 2017
Another Look at the Wage Growth Tracker's Cyclicality
Though Friday's employment report showed that payroll employment rose by a robust 222,000 jobs in June—much higher than most forecasts—enthusiasm for the news was tempered somewhat by average hourly wages coming in below expectations. Is the (ongoing) relatively tepid pace of wage growth a cause for concern? Perhaps, but the ups and downs of average wages over the course of the business cycle—the pattern of expansion-recession-expansion that typifies modern economies—are a bit more complicated than they may seem.
The year-over-the-year growth in the average wage level that we see in the official employment conditions report is influenced by wages paid to people who were employed either today or a year earlier. That is, the wages of those who remained employed (EE) as well as those who entered employment (NE) and those who exited employment (EN). Because the individuals in these groups may command different wages on average—due to experience, for example—the usual wage growth measures confound the effects of changes in the average wage of people with particular types of year-over-year employment histories. In that sense, the usual wage growth statistic may not exactly be comparing apples to apples.
Research by, for example, Solon, Barsky, and Parker 1992 and Daly and Hobjin 2016 explores the effect of the changing composition of workers over time using microdata on individuals with known employment histories. They show that people who enter and exit employment have a lower average wage than those who stay employed over the year and that the net exit/entry flow increases when the labor market is weak—more people leave employment, and fewer people enter it. As a result, the disproportionate increase in the net flow of workers with a lower-than-average wage serves to boost the overall average wage level during recessions.
One approach to making a more apples-to-apples comparison of average wages over time is to strip out the effect that comes from the change in the share of workers who stay employed and who entered or exited employment. Technically speaking, the composition-adjusted wage growth series is determined by adding the change in average log hourly wage within the EE group and the same change within the EN/NE group, while holding constant the respective average population shares in each group. The chart below illustrates the result of this adjustment.
I should note that the change in the average wage uses data only for people who have a known employment status a year earlier, which results in a wage growth series that is somewhat higher than the change in the average wage of all employed people, some of whom have an unknown employment history.
As the chart shows, relative to the adjusted series (the green line), growth in overall average wages (the orange line) stayed up longer during the last recession, then fell by less, and was slower to adjust to improving labor market conditions (falling unemployment) after the recession ended. The correlation between the overall growth in average wages and the inverse of the unemployment rate is 0.72, and this correlation rises to 0.79 using the adjusted wage growth series.
An alternative approach to making a more apples-to-apples comparison of average wages is to ignore the entry/exit margin and only look at people who are employed both today and a year earlier (EE). The Wage Growth Tracker (computed here as the difference in average log hourly wage) does that for the subset of EE people who have an actual wage record in both periods (no earnings information is collected for self-employed workers in the Current Population Survey). The following chart compares this version of the Wage Growth Tracker with the growth in overall average wages.
The Atlanta Fed's Wage Growth Tracker uses the median change in wages rather than the average change, but it displays very similar dynamics.
As the chart shows, the growth in average wages for those who remain in wage and salary jobs (the red line) is a bit smoother than growth in overall average wages (the orange line) and moves more in sync with the inverse of the unemployment rate (the correlation is 0.85). However, its level is quite a bit higher than growth in overall average wages. This disparity is because the average wage for those entering employment is less than for those exiting, so the change in average wages along the entry/exit margin is always negative.
But enough math—let's put this all together. If you want a measure of wage growth that reflects relative labor market strength, then looking at wage growth after controlling for entry/exit composition effects is probably a good idea. The Wage Growth Tracker seems to do that job reasonably well. However, the Wage Growth Tracker almost certainly overstates the growth in per hour wage costs that employers are facing. Most importantly, it ignores the employment exit/entry margin. Hence, one should avoid interpreting the Wage Growth Tracker as a direct measure of growth in labor costs—a point also discussed in this recent Atlanta Fed podcast episode . The next reading from the Wage Growth Tracker will be available when the Census Bureau releases the Current Population Survey microdata, usually within a couple of weeks of the national employment report. Given that the unemployment rate has remained relatively low recently, I would expect the Wage Growth Tracker to stay at a relatively high level. Check back here then and we'll see what we learn.
May 22, 2017
GDPNow's Second Quarter Forecast: Is It Too High?
Real gross domestic product (GDP) growth slowed from a 2 percent pace in 2016 to an annual rate of 0.7 percent in the first quarter of 2017. The Federal Open Market Committee viewed this slowdown in growth "as likely to be transitory," according to its last statement.
Indeed, current quarter GDP forecasting models maintained by the Federal Reserve Banks of New York, St. Louis, and Atlanta have been pointing toward stronger second quarter growth (2.3 percent, 2.6 percent and 4.1 percent, as reported on their respective websites on May 19, 2017).
The Atlanta Fed's model—GDPNow—is at the high end of this range and is also high relative to other professional forecasts. The median forecast for second quarter real GDP growth in the May Survey of Professional Forecasters (SPF) was 3.1 percent, for instance, and recent forecasts from Blue Chip Publication surveys displayed on our GDPNow page show some divergence from our model as well.
We encourage—and frequently receive—feedback on our GDPNow tool, and some users have suggested that our forecast for second quarter growth is too high. In fact, some empirical evidence supports that view. The evidence considered here correlates differences between consensus Blue Chip Economic Indicators Survey and GDPNow forecasts for growth about 80 days before the first GDP release with the GDPNow forecast errors (see the chart below).
A note about the chart: The horizontal axis shows the difference between the Blue Chip consensus forecasts and GDPNow's forecast. The vertical axis measures the 80-day-ahead GDPNow forecast error, defined as the difference between the first published estimate of real GDP growth and the GDPNow forecast at the time of the mid-quarter Blue Chip survey.
As the chart shows, there is a positive relationship between the Blue Chip-GDPNow discrepancy and the GDPNow forecast error. A simple linear regression would predict that the GDPNow forecast of 3.7 percent growth on May 5 was too high by nearly 1.0 percentage point. Moreover, the chart suggests that there has been a bias in GDPNow forecasts since the fourth quarter of 2015 of between 0.9 and 2.0 percentage points at the time of these mid-quarter Blue Chip surveys. If you are inclined to think the GDPNow forecast for second quarter growth is a bit too high, then this evidence will not change your mind.
Given this evidence, you might think that putting relatively little stock in the GDPNow forecast at this point in the quarter would be prudent. Indeed, if we calculate the weighted average of the historical Blue Chip consensus and GDPNow forecasts that produced the most accurate forecast of the first estimate of real GDP growth, then the optimal weight of the GDPNow forecast lies somewhere between 0.34 and 0.55 (see the chart below). The weight depends on the number of days until the first GDP release.
For example, the optimal weight of 0.55 on GDPNow about 54 days before the first GDP release means that 0.55 times the GDPNow forecast plus 0.45 times Blue Chip consensus survey forecast has been more accurate, on average, than any other weighted average of the two forecasts. The lowest weight on GDPNow corresponds to forecasts made about 83 days before the first GDP release—the time when GDPNow's bean-counting algorithms have the least amount of source data to work with.
A weighted average of the Blue Chip consensus and GDPNow forecasts at that time would put the GDP forecast about 0.6 to 0.7 percentage points below the current GDPNow forecast. However, the confidence bands around these estimates are wide, so the positive weight placed on GDPNow early in the quarter could just be the result of chance.
Let's cut to the chase—why, exactly, is the GDPNow forecast for second quarter GDP growth so high? The details of the GDPNow forecast provide some clues. We can compare the GDPNow forecasts of GDP components with those from the SPF. (The Blue Chip forecast does not provide detail on all the GDP components.) The following table translates the median SPF forecasts into contributions to second quarter real GDP growth. These contributions are shown alongside GDPNow's forecasted contributions as well as the average contributions to real GDP growth over the prior four quarters.
Clearly, more than half of the difference between the GDP growth forecasts from GDPNow and the SPF is due to inventories. For both forecasts, inventory investment also accounts for over half of the pickup in second quarter growth from the trailing four-quarter average.
A macroblog post I wrote last year showed that the growth-forecast contribution of mid-quarter inventory investment produced roughly equivalent accuracy in the SPF and GDPNow models, but it was much less accurate than the contribution forecasts of the other GDP components. Based on experience, we can't be confident that either forecast of inventory investment is likely to be very accurate or that one is likely to be much more accurate than another.
With very little hard data in hand for the second quarter for most of the GDP components—and for inventories in particular—we will continue to closely monitor if the data are as strong as GDPNow is anticipating or if they hew more closely to other forecasts. Check back with us to see.
May 11, 2017
Are Small Loans Hard to Find? Evidence from the Federal Reserve Banks' Small Business Survey
The Federal Reserve Banks recently released results from the nationwide 2016 Small Business Survey, which asks firms with 500 or fewer employees about business and financing conditions. One key finding is just how small the financing needs of many businesses are. One-fifth of small businesses that applied for financing in the prior 12 months were seeking $25,000 or less. A further 35 percent were seeking between $25,001 and $100,000.
The data also show that firms seeking relatively small amounts of financing (up to $100,000) receive a significantly smaller fraction of their funding than firms who applied for more than $250,000. Chart 1 shows the weighted average of the share of financing received by the amount the firm was seeking.
So what explains this variation in financing attainment across the amount requested? We've heard reports from small business owners that smaller loans are relatively more difficult to obtain, especially from traditional banks. One often-cited rationale is that the administrative burden associated with originating and managing a small loan is often just not worth the bank's time. However, this notion is not entirely consistent with data on the current holdings of small business loans on the balance sheets of banks. As of June 2015, loans of less than $100,000 made up about 92 percent of the number of business loans under $1 million.
So it seems originating a loan for less than $100,000 is not uncommon for a bank after all. So why, then, do business owners say that smaller loans are more difficult to get? Using data from the 2016 Small Business Survey, we can investigate the reason for this apparent disconnect.
Much can be explained by looking at the characteristics of those who borrow small amounts versus large amounts. Firms seeking $25,000 or less are more likely to be high credit risk and younger, have fewer employees, and have smaller revenues than firms applying for more than $250,000. The table below summarizes the differences:
Of particular importance is the credit risk associated with the firm. Controlling for differences in this factor, it turns out that smaller amounts of financing are not more difficult to obtain. Charts 2 and 3 show the weighted average share of financing received by amount sought for low credit risk firms and for middle to high credit risk firms separately.
As charts 2 and 3 demonstrate, low credit risk firms are able to obtain a similar share of the amount requested, regardless of how much they applied for. The same is true for higher risk firms. We also see that medium and high risk firms get less of their financing needs met than low credit risk firms that apply for similar amounts.
From this evidence, it seems that credit approval has more to do with the attributes of the firm than the amount of financing for which the firm applied. These results also highlight the potential importance of alternatives to traditional bank financing so that riskier entrepreneurs—including important contributors to the dynamism of the economy such as startups—have somewhere to turn. A later macroblog post will explore how low and high credit risk firms use financing differently, including where they apply and where they receive funding.
- Hitting a Cyclical High: The Wage Growth Premium from Changing Jobs
- Thoughts on a Long-Run Monetary Policy Framework, Part 4: Flexible Price-Level Targeting in the Big Picture
- Thoughts on a Long-Run Monetary Policy Framework, Part 3: An Example of Flexible Price-Level Targeting
- Thoughts on a Long-Run Monetary Policy Framework, Part 2: The Principle of Bounded Nominal Uncertainty
- Thoughts on a Long-Run Monetary Policy Framework: Framing the Question
- What Are Businesses Saying about Tax Reform Now?
- A First Look at Employment
- Weighting the Wage Growth Tracker
- GDPNow's Forecast: Why Did It Spike Recently?
- How Low Is the Unemployment Rate, Really?
- April 2018
- March 2018
- February 2018
- January 2018
- November 2017
- October 2017
- September 2017
- August 2017
- July 2017
- May 2017
- Business Cycles
- Business Inflation Expectations
- Capital and Investment
- Capital Markets
- Data Releases
- Economic conditions
- Economic Growth and Development
- Exchange Rates and the Dollar
- Fed Funds Futures
- Federal Debt and Deficits
- Federal Reserve and Monetary Policy
- Financial System
- Fiscal Policy
- Health Care
- Inflation Expectations
- Interest Rates
- Labor Markets
- Latin America/South America
- Monetary Policy
- Money Markets
- Real Estate
- Saving, Capital, and Investment
- Small Business
- Social Security
- This, That, and the Other
- Trade Deficit
- Wage Growth