very important change of tone and interesting piece from gs.
September Payrolls—Why We Marked Down Our Estimate (McKelvey)
· Earlier today we downgraded our estimate for September’s change in nonfarm payrolls, to -250,000 from -200,000 previously. In this comment, we explain this decision in more detail. The report is due for release tomorrow morning at 8:30 Eastern time. (yah right!)
· Although many of the same factors would argue for a larger increase in the unemployment rate than the 0.1-point increase (to 9.8%) we have been anticipating, we have not changed this estimate. We do, however, see upside risk for this key indicator.
· In other parts of the report, we expect the Bureau of Labor Statistics to announce a preliminary estimate of about -150,000 to -200,000 for the next annual benchmark revision to nonfarm payrolls, and we also think hourly wages rose only 0.1% in September.
Although the US economy appears to have grown at about a 3% annual rate in the quarter that ended yesterday, the US labor market continues to shed jobs. Earlier today we downgraded our estimate for September’s change in nonfarm payrolls to -250,000 from -200,000 previously. Meanwhile, the median forecast as measured by Bloomberg has improved slightly, to -175,000 from -180,000. As a result, we are now well below “consensus.”
Our decision was based on several pieces of information that have come to light since we published our original estimate last Friday. In chronological order:
1. US consumers perceived a setback in job availability in September. In its monthly survey of confidence, the Conference Board asks consumers whether jobs are “plentiful,” “not so plentiful,” or “hard to get.” We have found that the difference between the percentages of the respondents choosing the extremes of this question—plentiful versus hard to get—helps modestly in explaining monthly changes in nonfarm payrolls. In September, this indicator widened to -43.6 (percent of the respondent sample) from -40.0 in August. The September reading was close to July’s -44.8, which currently stands as the lowest point for the current business cycle.
2. Monster, Inc.’s index of on-line advertising slipped in September. The decline was not large—only two index points—but it represents a small setback in what had been a modest trend of improvement relative to seasonal norms. Because the index is not seasonally adjusted (it goes back only to October 2003), this is most easily seen in the recent behavior of year-to-year changes in the index. In September, the index was -25.6% below its year-earlier level. This followed a progressive improvement from a low of -31.0% in April to -23.9% in August. The setback is obviously small compared to the improvement that preceded it, but it comports with the message in the Conference Board survey that job availability may have stalled last month. The Conference Board’s more comprehensive roundup of on-line job advertising sent a similar message.
3. The number of unemployed workers receiving regular and extended benefits continues to rise. Although continuing claims rose only slightly (17,000) between the reference weeks for the August and September payroll surveys, the number receiving extended benefits rose about 230,000 over the same period. (The reference weeks include the 12th of the month.) The two figures are not directly comparable, as the first is seasonally adjusted while the second cannot be given the occasional nature of the extended benefit programs. However, given that the increase in extended benefits occurred at a time when continuing benefits for regular programs normally decline (this figure was down nearly 460,000 before seasonal adjustment), we see the increase as strong evidence that the modest downtrend in continuing claims reflects mainly the exhaustion of regular benefits rather than a pickup in net hiring, especially given the foregoing observations on job availability. Thus, we do not take much comfort in the 70,000 drop in continuing claims reported for the week ending September 19, which in any event was after the payroll survey was taken, and the stickiness in initial claims likewise suggests that the US labor market remains challenging.
4. The Institute for Supply Management’s (ISM’s) index of employment in manufacturing also stalled in September. After rising sharply from a record low (26.1) in February, this index dipped 0.2 points to 46.2. Thus, we see no reason to expect a meaningful change from the loss of 63,000 jobs reported for the manufacturing sector in September. Job losses in manufacturing are hardly news; in recent years the ISM employment index would need to climb into the mid 50s just to signal stabilization in this component of the payroll report. However, given that manufacturing is currently the bright spot in the US growth outlook, a stalling in the employment index for this sector suggests a lack of progress elsewhere in the US labor market. After all, the index does have some explanatory power for nonfarm payrolls, especially if its cousin for the nonmanufacturing survey is unavailable, as it is in this case.
5. More generally, economic indicators for August and September have been decidedly mixed. Or should we say undecidedly mixed? For example, today’s ISM composite index also suggested that the pickup in manufacturing activity paused, and while construction outlays rose a bit more than expected in August the level of activity in this sector turns out to have been quite a bit lower than previously thought. Earlier, orders and shipments for durable goods, including those for capital goods, were also reported to have fallen much more than expected in August.
Although we do not usually forecast job changes by sector, it is worth directing some attention to a potential setback in state and local payrolls in tomorrow’s report. September is the month when returning teachers (and other educational workers) show up in the payroll data. Although most jurisdictions have tried their best to avoid outright layoffs in this area, they have also clamped down on new hires. This suggests that we could see job losses on a seasonally adjusted basis, as jobs in this sector before seasonal adjustment normally rise about 14%, or roughly 290,000 at current levels of the workforce, between August and September.
In recent years, the September labor market report has also included a preliminary estimate of the annual benchmark revision that the Bureau of Labor Statistics (BLS) will implement early next year. This revision lines up the monthly payroll figures, which are based on sample data, with the Quarterly Census of Employment and Wages (QCEW), a near-universe data set base on employers’ payments into the state unemployment insurance funds. Specifically, the upcoming benchmark will align the data for March 2009, replacing the add factor based on the infamous “birth/death” model that estimates the effects of net business formation on the job count. Based on our seasonal adjustment of QCEW data through December 2008 (March figures are not yet available), we expect the BLS to estimate that this revision will be on the order of -150,000 to -200,000. Apart from the announcement effect, this should have no bearing on other parts of the report until the BLS actually implements the revision with next February’s data. At that time, the BLS may also decide to shave the post March 2009 add-ons from the birth/death model, which have seemed large given the extremely low level of economic activity.
Many of the points raised earlier may also be signaling a larger-than-expected jump in the jobless rate. However, our estimate for this key indicator remains 9.8%, a 0.1-point uptick from the figure reported for August, for two reasons. First, the 50,000 change in our payroll forecast is worth “only” about 3 basis points (0.03 point) on the unemployment rate. Second, the unrounded rate in August was low, at 9.657%. To round up to 9.9%, the unrounded rate would thus have to rise by more than 19 basis points—a net addition of almost 300,000 to the ranks of the unemployed.
One big question in this regard is whether the BLS has underestimated the size of the labor force —the number of people who are employed plus those who are without jobs but actively looking for work—in recent months. Since an individual’s employment status is much easier to establish than the degree of effort he or she has been applying to looking for a job, errors in measuring the labor force will be highly correlated with the number of people classified as unemployed, the numerator of the unemployment rate. Thus, if the labor force has been understated and corrects to a more normal level, then the unemployment rate will tend to rise.
The case for underestimation comes from the labor force figures reported for recent months. In August, the BLS reported a net increase of 73,000. However, this recovered only a small part of the 577,000 cumulative decline reported for June and July. Since the labor force rarely trends down for long, even during recessions, this would suggest upside risk to the unemployment rate over the next few months, especially if the economy is crawling back onto a recovery path and individuals perceive an improvement—however faint—in labor market conditions.
The counterargument is based on a longer history of the labor force data. Over the past year, for example, the labor force has been essentially flat—down 0.16%, or a net loss of about 250,000 workers. Although year-to-year declines are quite rare, so is the depth of this recession and, by extension, its influence on workers’ perceptions about job availability. In this view of the world, the role of job market attitudes in our forecast adjustment for payrolls cautions against a parallel increase in our estimate of the unemployment rate.
Beyond the changes in nonfarm payrolls and the unemployment rate, we will look closely at hourly earnings and the index of hours worked. With the latest minimum wage hike already in the August data, we expect the weakening trend in hourly wages to reemerge. Thus, we estimate only a 0.1% increase in this indicator. Meanwhile, our estimate for nonfarm payrolls—down about 0.2%—leaves little room for a significant increase in hours worked unless workweeks lengthened considerably. With the data turning a bit squishy, this seems unlikely. Accordingly, this payroll report is unlikely to create much enthusiasm for strong income gains (the byproduct of hours and wages) or production, though in the latter case substantial productivity gains can go a long way in filling the gap between anemic hours and moderate to firm growth in real GDP.
Ed McKelvey
· Earlier today we downgraded our estimate for September’s change in nonfarm payrolls, to -250,000 from -200,000 previously. In this comment, we explain this decision in more detail. The report is due for release tomorrow morning at 8:30 Eastern time. (yah right!)
· Although many of the same factors would argue for a larger increase in the unemployment rate than the 0.1-point increase (to 9.8%) we have been anticipating, we have not changed this estimate. We do, however, see upside risk for this key indicator.
· In other parts of the report, we expect the Bureau of Labor Statistics to announce a preliminary estimate of about -150,000 to -200,000 for the next annual benchmark revision to nonfarm payrolls, and we also think hourly wages rose only 0.1% in September.
Although the US economy appears to have grown at about a 3% annual rate in the quarter that ended yesterday, the US labor market continues to shed jobs. Earlier today we downgraded our estimate for September’s change in nonfarm payrolls to -250,000 from -200,000 previously. Meanwhile, the median forecast as measured by Bloomberg has improved slightly, to -175,000 from -180,000. As a result, we are now well below “consensus.”
Our decision was based on several pieces of information that have come to light since we published our original estimate last Friday. In chronological order:
1. US consumers perceived a setback in job availability in September. In its monthly survey of confidence, the Conference Board asks consumers whether jobs are “plentiful,” “not so plentiful,” or “hard to get.” We have found that the difference between the percentages of the respondents choosing the extremes of this question—plentiful versus hard to get—helps modestly in explaining monthly changes in nonfarm payrolls. In September, this indicator widened to -43.6 (percent of the respondent sample) from -40.0 in August. The September reading was close to July’s -44.8, which currently stands as the lowest point for the current business cycle.
2. Monster, Inc.’s index of on-line advertising slipped in September. The decline was not large—only two index points—but it represents a small setback in what had been a modest trend of improvement relative to seasonal norms. Because the index is not seasonally adjusted (it goes back only to October 2003), this is most easily seen in the recent behavior of year-to-year changes in the index. In September, the index was -25.6% below its year-earlier level. This followed a progressive improvement from a low of -31.0% in April to -23.9% in August. The setback is obviously small compared to the improvement that preceded it, but it comports with the message in the Conference Board survey that job availability may have stalled last month. The Conference Board’s more comprehensive roundup of on-line job advertising sent a similar message.
3. The number of unemployed workers receiving regular and extended benefits continues to rise. Although continuing claims rose only slightly (17,000) between the reference weeks for the August and September payroll surveys, the number receiving extended benefits rose about 230,000 over the same period. (The reference weeks include the 12th of the month.) The two figures are not directly comparable, as the first is seasonally adjusted while the second cannot be given the occasional nature of the extended benefit programs. However, given that the increase in extended benefits occurred at a time when continuing benefits for regular programs normally decline (this figure was down nearly 460,000 before seasonal adjustment), we see the increase as strong evidence that the modest downtrend in continuing claims reflects mainly the exhaustion of regular benefits rather than a pickup in net hiring, especially given the foregoing observations on job availability. Thus, we do not take much comfort in the 70,000 drop in continuing claims reported for the week ending September 19, which in any event was after the payroll survey was taken, and the stickiness in initial claims likewise suggests that the US labor market remains challenging.
4. The Institute for Supply Management’s (ISM’s) index of employment in manufacturing also stalled in September. After rising sharply from a record low (26.1) in February, this index dipped 0.2 points to 46.2. Thus, we see no reason to expect a meaningful change from the loss of 63,000 jobs reported for the manufacturing sector in September. Job losses in manufacturing are hardly news; in recent years the ISM employment index would need to climb into the mid 50s just to signal stabilization in this component of the payroll report. However, given that manufacturing is currently the bright spot in the US growth outlook, a stalling in the employment index for this sector suggests a lack of progress elsewhere in the US labor market. After all, the index does have some explanatory power for nonfarm payrolls, especially if its cousin for the nonmanufacturing survey is unavailable, as it is in this case.
5. More generally, economic indicators for August and September have been decidedly mixed. Or should we say undecidedly mixed? For example, today’s ISM composite index also suggested that the pickup in manufacturing activity paused, and while construction outlays rose a bit more than expected in August the level of activity in this sector turns out to have been quite a bit lower than previously thought. Earlier, orders and shipments for durable goods, including those for capital goods, were also reported to have fallen much more than expected in August.
Although we do not usually forecast job changes by sector, it is worth directing some attention to a potential setback in state and local payrolls in tomorrow’s report. September is the month when returning teachers (and other educational workers) show up in the payroll data. Although most jurisdictions have tried their best to avoid outright layoffs in this area, they have also clamped down on new hires. This suggests that we could see job losses on a seasonally adjusted basis, as jobs in this sector before seasonal adjustment normally rise about 14%, or roughly 290,000 at current levels of the workforce, between August and September.
In recent years, the September labor market report has also included a preliminary estimate of the annual benchmark revision that the Bureau of Labor Statistics (BLS) will implement early next year. This revision lines up the monthly payroll figures, which are based on sample data, with the Quarterly Census of Employment and Wages (QCEW), a near-universe data set base on employers’ payments into the state unemployment insurance funds. Specifically, the upcoming benchmark will align the data for March 2009, replacing the add factor based on the infamous “birth/death” model that estimates the effects of net business formation on the job count. Based on our seasonal adjustment of QCEW data through December 2008 (March figures are not yet available), we expect the BLS to estimate that this revision will be on the order of -150,000 to -200,000. Apart from the announcement effect, this should have no bearing on other parts of the report until the BLS actually implements the revision with next February’s data. At that time, the BLS may also decide to shave the post March 2009 add-ons from the birth/death model, which have seemed large given the extremely low level of economic activity.
Many of the points raised earlier may also be signaling a larger-than-expected jump in the jobless rate. However, our estimate for this key indicator remains 9.8%, a 0.1-point uptick from the figure reported for August, for two reasons. First, the 50,000 change in our payroll forecast is worth “only” about 3 basis points (0.03 point) on the unemployment rate. Second, the unrounded rate in August was low, at 9.657%. To round up to 9.9%, the unrounded rate would thus have to rise by more than 19 basis points—a net addition of almost 300,000 to the ranks of the unemployed.
One big question in this regard is whether the BLS has underestimated the size of the labor force —the number of people who are employed plus those who are without jobs but actively looking for work—in recent months. Since an individual’s employment status is much easier to establish than the degree of effort he or she has been applying to looking for a job, errors in measuring the labor force will be highly correlated with the number of people classified as unemployed, the numerator of the unemployment rate. Thus, if the labor force has been understated and corrects to a more normal level, then the unemployment rate will tend to rise.
The case for underestimation comes from the labor force figures reported for recent months. In August, the BLS reported a net increase of 73,000. However, this recovered only a small part of the 577,000 cumulative decline reported for June and July. Since the labor force rarely trends down for long, even during recessions, this would suggest upside risk to the unemployment rate over the next few months, especially if the economy is crawling back onto a recovery path and individuals perceive an improvement—however faint—in labor market conditions.
The counterargument is based on a longer history of the labor force data. Over the past year, for example, the labor force has been essentially flat—down 0.16%, or a net loss of about 250,000 workers. Although year-to-year declines are quite rare, so is the depth of this recession and, by extension, its influence on workers’ perceptions about job availability. In this view of the world, the role of job market attitudes in our forecast adjustment for payrolls cautions against a parallel increase in our estimate of the unemployment rate.
Beyond the changes in nonfarm payrolls and the unemployment rate, we will look closely at hourly earnings and the index of hours worked. With the latest minimum wage hike already in the August data, we expect the weakening trend in hourly wages to reemerge. Thus, we estimate only a 0.1% increase in this indicator. Meanwhile, our estimate for nonfarm payrolls—down about 0.2%—leaves little room for a significant increase in hours worked unless workweeks lengthened considerably. With the data turning a bit squishy, this seems unlikely. Accordingly, this payroll report is unlikely to create much enthusiasm for strong income gains (the byproduct of hours and wages) or production, though in the latter case substantial productivity gains can go a long way in filling the gap between anemic hours and moderate to firm growth in real GDP.
Ed McKelvey
http://www.google.com/finance?q=GOOGLEINDEX_US:JOBS
回覆刪除google 的工具不少...看看他對非農等判斷會不會比較準阿?!
這是在回應上一篇的:
回覆刪除In short, no one ever got fired for owning Treasuries !?
沒有阿,就是GS自家的看法.我猜McKelvey應該鬆了一口氣.
回覆刪除