 | |||||||||||
| |||||||||||
|
« No Panacea | Main | Former Democrat Aide Allegedly Spied for Saddam » March 11, 2004Think Tank: Household Survey Accurate, Employment SurgesThe Heritage Foundation has published a new study that explores the difference between the federal government's Employer Survey, which shows stagnant job growth, with the Household Survey, which shows a surge in employment.
The payroll survey may be systematically undercounting job growth, creating an unprecedented job growth gap between its total employment measure and the household survey's. In the past six months, the BLS has approved new techniques to smooth the household survey's measure of total employment in order to make month-to-month comparisons. Analysts can now point with confidence to the employment of a record number of Americans as of January 2004 and the employment of an additional 2.2 million workers since the recession ended. [Via Bruce Bartlett] By the way, you can't expect the press will take the time to read and digest the Heritage study - they will continue to pound the data from the Employer Survey (for as long as it shows bad news) and ignore the Household Survey, in order to portray the Bush Boom as a failure. It's your job to arm yourself with the truth and tell it to as many people you can, and encourage them to tell others. Posted in Economy & Business
| Linked By |
Please support HobbsOnline by doing your online shopping at Amazon.com Comments
2003 employment numbers in South Carolina are already being revised upward, according to today's The State (Columbia, SC).I found this article last week that says that the employer survey is usually revised upward during a recovery. Apparently, it's true. But the Democrats will never admit it. Posted by: Mike D in SC at March 11, 2004 02:40 PMDr. Kane fails to make his case. We can argue, back and forth, as to whether or not he is persuasive but, in the final analysis, it isn't even in question as to which survey economists and Wall Street place their faith in. Some observers have speculated that the household survey provides a better indication of the trend in employment at and around turning points in the business cycle. It is our judgment that the payroll survey provides more reliable information on the current trend in wage and salary employment. The payroll survey has a larger sample than the household survey--400,000 business establishments covering about one-third of The CBO uses the payroll survey. Wall Street uses the payroll survey. Dr. Kane would be well-served to address why--if the household survey is more accurate--Wall Street (which lives and dies on the timeliness and accuracy of data) wouldn't use it.
JadeGold - long after Galileo proved the earth orbited the sun, lots of learned people insisted the sun orbited the earth. They weren't right - but they hadn't yet been convinced otherwise. Dr. Kane's analysis was released 7 days ago. Even if he is 1,000 percent right, Wall Street and the CBO aren't going to shift on a dime. Be realistic. Wall Street, etc., have spent years using the payroll survey and it will take a lot more than one study to convince them to shift. That doesn't mean Kane is wrong. It just means that the process of re-evaluating the accuracy of and problems with each survey is now under way, and the process of re-educating Wall Street and the CBO has just begun. Rather than ask Dr. Kane why Wall Street hasn't been convined yet, a mere one week after his study was released, why don't you print Dr. Kane's study, pore over it word by word, apply your Ph.D. in economics (if you have one) and either confirm or refute it based on facts and reasonable analysis. Your comment above amounts to saying Dr. Kane is wrong because some other folks don't agree with him. Um, okay, let's apply that illogic to you. You say Bush is bad. I and some other folks don't agree with you. So, based on your logic, you must be wrong - Bush is NOT bad. Great! Now that, using your logic, you are convinved that Bush is good, i know you'll be voting for Bush! Posted by: Bill Hobbs at March 11, 2004 04:11 PM[Bill] Why do you or anyone else bother arguing with JadeGold? As I showed back in the discussion, eons ago, about Bush being "AWOL," that [b]JadeGold does not respond.[/b] He doesn't answer questions. He doesn't consider arguments. He won't engage. He often doesn't even [i]pretend[/i] to answer the call of the question. He reads from a script, like a robot. Arguing with JadeGold is like trying to debate your answering machine. I asked him a simple question: I asked if he knew whether or not it was customary in the Alabama Air National Guard for pilots to miss monthly meetings and make them up later. This question goes to the core of the charge... because it if were normal, then there is nothing wrong with Bush doing the same. It's not an insulting question. It's not a misleading question. If he didn't know, he could simply have said "Dafydd, I don't know." But he didn't. He didn't respond. I asked him about fifteen times, and he just ignored me -- and my evidently devastating question -- for day after day after day, while he was pretending to argue with others. He couldn't answer the question (or he suspected the answer wouldn't do his side any good), so he just pretended the question had never been asked. Even when others took up my question and also asked JadeGold, he ignored them, too. So why, I ask again, should anyone in this blog-comment section bother responding to JadeGold about [i]anything?[/i] If he has so much contempt for us and for the basic rules of polite discourse, then he is intellectually dead... and we should do him the courtesy of treating him as such. Dafydd [Bill] (Repost so I can use all the cool HTML code stuff by replacing [ ] with angle brackets...!) Why do you or anyone else bother arguing with JadeGold? As I showed back in the discussion, eons ago, about Bush being "AWOL," JadeGold does not respond. He doesn't answer questions. He doesn't consider arguments. He won't engage. He often doesn't even pretend to answer the call of the question. He reads from a script, like a robot. Arguing with JadeGold is like trying to debate your answering machine. I asked him a simple question: I asked if he knew whether or not it was customary in the Alabama Air National Guard in the early 1970s for pilots to miss monthly meetings and make them up later. This question goes to the core of the charge... because it if were normal, then there is nothing wrong with Bush doing it. It's not an insulting question. It's not a misleading question. If he didn't know, he could simply have said "Dafydd, I don't know." But he didn't. He didn't respond. I asked him about fifteen times, and he just ignored me -- and my evidently devastating question -- for day after day after day, while he was pretending to argue with others. He couldn't answer the question (or he suspected the answer wouldn't do his side any good), so he just made believe the question had never been asked. Even when others took up my question and also asked JadeGold, he ignored them, too. So why, I ask again, should anyone in this blog-comment section bother responding to JadeGold about anything? If he has so much contempt for us and for the basic rules of polite discourse, then he is intellectually dead... and we should do him the courtesy of treating him as such. Dafydd Posted by: Dafydd ab Hugh at March 11, 2004 04:42 PMI think Jade needs to be answered. The well crafted rebuttals he posts make you question what Bill has posted. In so doing, you become better informed about the subject and better able to refute idiotarian arguements. Posted by: Matt S. at March 11, 2004 04:45 PMWhy don't you go to BLS website and read this recent notice on the divergence of HH and payroll surveys? http://www.bls.gov/cps/ces_cps_trends.pdf Possible causes of differences in employment trends The following summarizes some issues with the surveys that are important when comparing changes in employment from the two sources. 1. Sampling error – The payroll survey has a much larger sample size than the household survey. The payroll survey’s active sample covers approximately 400,000 business establishments of all sizes representing about one-third of total nonfarm employment. The household survey is much smaller at 60,000 households, covering a very small fraction of total employed persons. Household survey employment is therefore subject to larger sampling error, more than double that of the payroll survey on a monthly basis (see Box 1). When looking at short-term trends in either survey, especially over-the-month changes, it is essential to assess the statistical significance of the change. When comparing the two series over longer periods of time, however, other factors also need to be considered; some of these are discussed below. 2. Payroll survey benchmark – The payroll survey estimates are benchmarked once a year against a full universe count of employment derived from Unemployment Insurance tax records that nearly all employers are required to file. The payroll survey’s most recent benchmark—to March 2003 employment records—resulted in a small downward revision of one-tenth of one percent, indicating that the survey estimates were accurately tracking the universe of nonfarm payroll employment (see additional information on the payroll survey benchmark below). The household survey does not have a benchmark source for employment. New business births in the payroll survey – The payroll survey sample does not include new firms immediately. They are incorporated with a lag. In the interim, a model-based estimate is used each month to account for employment resulting from new firm births. Based on the relatively small benchmark revision for March 2003, as well as comparisons to universe data for Second Quarter 2003, the model appears to be performing well during the recent period. 3. Population controls in the household survey – Population controls determine the weights used in the household survey to adjust the sample results to the overall level of the U.S. population. The population controls are developed by the U.S. Census Bureau. They are derived from decennial census information and, between census years, from administrative and other data. There are limitations to the population control estimates due primarily to the difficulties associated with estimating the net international migration component. The population controls contributed significantly to the discrepancy between payroll and household survey employment in the 1980s and 1990s when the household survey showed less growth than the payroll survey. The upward trend in household employment since the end of the 2001 recession has been largely a function of the estimated growth in population. That is to say, the household survey has not shown an increase in the proportion of the population that is employed. In fact, the employment-population ratio has declined since 2001: it was 64.3 percent at the start of the recession (March 2001) and 63.0 percent at the trough. The ratio declined further from late 2002 to 2003 and has been fairly flat over the past year. The increase in the household survey employment level, therefore, has been driven by the estimated growth in the population. (See additional information on the recent household survey population control adjustments below.) Posted by: bagh at March 11, 2004 05:39 PM "A brief introduction to statistics would confirm that a survey sample of 60,000 is more than adequate. Much smaller samples are used every day in media polls assessing everything from voter attitudes to consumer product ratings, and they are perfectly valid." According to BLS sample size does matter: http://www.bls.gov/cps/ces_cps_trends.pdf Size of over-the-month change in employment required for a statistically significant movement: Tim Kane: "The result of benchmarking is another lesson in caution, as post-benchmark CES data are usually—not rarely—quite different from the pre-benchmark series. Benchmark revisions for March 2003 did not yield major changes. BLS concluded that the model appears to be performing well during the recent period. Posted by: bagh at March 11, 2004 06:02 PMAlan Greenspan: " "I wish I could say the household survey were the more accurate. Everything we've looked at suggests that the data on so-called payrolls survey is surely the most accurate of the two and our suspicion is that at the end of the day there will be revisions to the household data." To test the self-employment theory, the Fed adjusted the household survey by taking out all the kinds of workers who do not show up on the payroll survey - self-employed people, but also farm workers and family workers in family-run companies. Even then, Mr. Greenspan said, the discrepancy remains large. The Fed's conclusion was that the household survey's results have been inflated by overestimates of population growth. Because the household survey is a sample, the Bureau of Labor Statistics infers the total change in jobs by multiplying the ratio of employed to unemployed workers in the household survey by its estimate of the total population. If the population estimate is too high, the estimated number of jobs will also be too high. The Bureau of Labor Statistics bases its population estimate on the 2000 census, but it updates that estimate yearly with data on births, deaths and immigration. Immigration numbers are largely guesswork, however, because so much immigration is illegal. Fed officials suspect that the immigration estimate is inflated, because it fails to reflect tighter immigration controls after Sept. 11, 2001, as well as declines caused by the economic slowdown... http://www.nytimes.com/2004/02/22/business/yourmoney/22view.html?pagewanted=print&position= Tim Kane: Average unemployment duration is 19.6 weeks". This is Dec2003 figure. Jan 2004: 19.8 weeks According to Labor Department data going back to 1948, the longest spell was 20.8 weeks in 1983. Bagh...why don't you address the problems with the PAYROLL survey as pointed out in the linked article? Or didn't you read it? Posted by: Syl at March 11, 2004 09:20 PMI certainly consider the Household Survey to be accurate, and I don't need that many stats to do it. The Labor Dept last Friday showed a labor force drop of about 265,000 in the Houshhold Survey (to now 146.5 million). Anyone who has ever seen the tourism industry (!) up close is not surprised by how much seasonal work there is. The Household Survey tends to reflect how much seasonal work is done in the nation, something no Payroll Survey has yet to accurately reflect. Posted by: Brad S at March 12, 2004 10:52 AMOpen your eyes and read it: The upward trend in household employment since the end of the 2001 recession has been largely a function of the estimated growth in population. That is to say, the household survey has not shown an increase in the proportion of the population that is employed. In fact, the employment-population ratio has declined since 2001: it was 64.3 percent at the start of the recession (March 2001) and 63.0 percent at the trough. The ratio declined further from late 2002 to 2003 and has been fairly flat over the past year. The increase in the household survey employment level, therefore, has been driven by the estimated growth in the population. Plain and simple... If there were a job growth, the employment-population ratio would be higher. It is not: Last month's participation rate of 65.9% was the lowest since September 1988. The average unemployment spell is now lasting more than four-and-a-half months (20.3 weeks), just slightly below the highest level in two decades—20.4 weeks—posted in January 1984, when the unemployment rate was 8%. The 1.6% annual growth rate of the hourly wages of blue-collar workers in manufacturing and non-managers in services is tied with the lowest rate on record (posted in December 1986), going back to 1964. Even with low inflation rates running at about 2%, such slow wage growth means many workers are falling behind in real terms. If Bush is so confident of job growth in his administration, why don't we see campaign ads emphasizing the number of jobs we have gained in the last 3 years????? Posted by: bagh at March 12, 2004 01:10 PM[Bagh] You have mentioned several problems with the household survey. Some could be important, such as that job growth may be in part spurred by population growth; others are just plain silly, such as pretending that a survey of 60,000 households cannot properly reflect a population of about a hundred million (polls of a mere 1500 are regularly used to reflect a voting population of 180 million with a margin of error of only 4%). So let's examine some of the problems with the establishment survey. Can you please tell me whether the establishment (payroll) survey counts independent contractors who work in a company, side by side with employees, one cube over, with the same boss, on the same projects -- but who invoice the company and are paid from the vendor account, rather than from the payroll account? And does the payroll survey count the self-employed -- people who own single-person or two-person businesses, small family businesses, and so forth? And does the payroll survey count people who work for small, start-up companies? If you admit that the payroll survey does not count contractors, new small business workers, and the self-employed, then can you estimate how many such workers there were in January 2001, how many there are today, and how many used to work for regular companies, hence were picked up the establishment survey... and who thus would falsely appear to have become "unemployed" according to that survey? Thanks, Dafydd ab Hugh Posted by: Dafydd ab Hugh at March 12, 2004 02:30 PMTim Kane: Average unemployment duration is 19.6 weeks".
Bagh, are you aware that there is a positive correlation between the length of unemployment benefits and the average length of unemployment? I don't remember who did the study, but they showed that as unemployment benefits are extended, average unemployment time increases. We've been extending unemployment benefits over the past 2 years; surely that has had an effect on the average length of unemployment. Yet another economist--this time a very conservative and pro-Bush--trashes the household survey. Robert Barro in the WSJ: One tempting hypothesis is that the household numbers look better because of a large expansion of the self-employed -- illustrated perhaps by the growth of full-time retailers on eBay.
Another hypothesis that does not work is that the payroll survey substantially undercounts the recent growth of employment in new firms. When the BLS used unemployment-insurance records through March 2003 to update the payroll survey for new-firm growth, this adjustment did not help to explain the mysterious gap between the household and payroll numbers. Really, there is no argument; the payroll survey is the standard. Arguing the household survey is actually more accurate is a bit like trying to argue the Detroit Tigers have a better team than the Yankees. [Folks] It is patently obvious that neither Bagh nor the execrable JadeGold has even bothered to read the report by the Heritage Foundation referenced in the original blogpost that began this discussion. Every objection they have raised is not only anticipated but factored into the original report. First, Bagh's obsessive harping on the fact that the household survey is not corrected for population changes is complete nonsense, because the BLS has recently done exactly that, releasing new data series that do, in fact, smooth the month-by-month reports of the household survey for population shifts. It makes no difference; the gap between the two surveys persists, even when population changes are taken into account. Next, his fretting that the household survey is wanting because it surveys "only" 60,000 households (out of 100 million), while the payroll survey gathers data from 400,000 employers, is utter nonsense, as the most basic statistical understanding should have told him. I can only conclude, if Bagh is being honest, that he simply doesn't have even that cursory level of understanding of statistics. The precision gained by shifting from a survey of 60,000 to one of 400,000 is so minimal as to vanish completely in the baseline uncertainty of conducting any survey at all. It could not possibly account for a gap of three million jobs. Not only that, but if Bagh is relying on the statistical error of the payroll survey -- then maybe he could explain why that error is so persistently in the same direction, that of supposedly overcounting jobs? If you go to a restaurant where they frequently add up your check wrong, but for some mystery reason the error is always in favor of the restaurant, then you can be pretty darned sure it's not due to random math errors! Purely statistical errors cannot possibly explain either the size of the gap or its persistent arrow pointing from a low payroll survey to a high household survey. There must be a systemic error in one or both surveys. But which one? Easily resolved: the one that matches the other economic indices -- rapidly rising GDP, falling unemployment claims, falling unemployment rate, rising tax receipts, and business investment -- is almost certainly the one that is accurate. That one is the household survey... not the payroll survey. There are also several problems with the payroll survey that Bagh and the coward briefly touch upon, but for which they show clear indication of not having read the report: first, the idea that self-employment cannot explain the gap because the household survey counted "only" 650,000 newly self-employed people, only 20% of the gap. The problem with this is that it's just the tip of the iceberg: economists have already noted that a lot of people who are counted as "self employed" by the payroll survey -- that is, who are not counted in the non-farm payroll reporting, hence not picked up at all by that survey -- do not call themselves "self employed" on the household survey. This is the group I identified months ago (I'm not an economist) as the contractor class; it jumped out at me immediately because I've been an independent contractor most of my life. We don't get paid out of a payroll account; I send an invoice to my "employer," and they pay out of their vendor account, just like they would pay the guys who come to fix the fax machine. But if asked at the time, I would not have called myself "self employed": I got up and drove to work, sat in a cubicle right between regular employees, I had the same boss, worked on the same contracts, went to the company picnics, and often got to participate in the employee stock-option plan. Solid estimates of this group indicates it's likely as big as (perhaps bigger than) the group of self-employed people who actually call themselves "self employed" on the household survey... which means we're not talking about "only" 650,000 new non-counted workers but a whopping 1.3 million at a minimum. Second, the payroll survey has a history of massive revisions, often years after the original publication. Merely checking back a couple of months later, or even the year-end revision, doesn't even begin to fix the problems. These revisions are most noticible during the recovery from a recession. For example, the last time we were recovering from a recession, 1991-1992, the BLS initially claimed that during this period, 800,000 net jobs were lost; but today, their figures for the same period show 600,000 net jobs were actually gained! That is a difference of 1.4 million jobs "created" (ex post facto) simply by the BLS's own revisions. Finally, neither Bagh nor the coward even stoops to comment on the most serious payroll-survey problem discussed by the report... almost certainly because neither of them managed to make it all the way to the end, if they even started reading. A case of MEGO, perhaps. (But one also wonders what level of math either of the pair attained at university.) And that is that during boom times, such as the mid to late 1990s, the payroll survey tends to overcount jobs due to higher turnover rates. If an employee leaves company A and is immediately re-employed by company B -- for example, if a headhunter snags him with a better offer -- then he will almost certainly appear on the payroll records of both companies during that period. In other words, Bill Hobbs is working for IBM until March 8th, when he leaves IBM and begins working for Cisco Systems on the 10th. But for that weekly reporting period, Bill is listed as having two jobs, working simultaneously for both IBM and Cisco. If turnover were constant, this wouldn't be a problem; the same artificially high report wouldn't show any net change due to turnover. But the problem comes when we transition from a boom to a recession to a recovery. Analysis from many independent economists, as well as from the BLS itself, shows that during the recovery from a recession, turnover rates drop markedly. The reason for this should be clear: people are more nervous about employment, so they tend to value stability and security over a slight increase in salary. (In today's case, the added insecurity of the 9/11 attacks and two wars add to this effect.) Since turnover leads to overcounting, a significant drop in turnover leads to a drop in overcounting... which looks on paper like a drop in jobs, despite the fact that, by definition, no job has been lost... instead of jumping from IBM to Cisco, Bill just decides to stay at IBM because he's worried. He's still employed; he is just no longer being falsely counted as doubly-employed. Using careful estimates from these economists, the authors of the Heritage Foundation report discovered that in fact, decreasing turnover alone accounts for between 1.0 million and 1.3 million job "losses" on the payroll survey. So let's add it up: a minimum of 1.3 million jobs "lost" to self employment and contract employment; a minimum of 1.0 million jobs "lost" because they never existed in the first place, but were the result of turnover-induced overcounting; and an unknown number of jobs "lost" due to the preliminary nature of all payroll data combine to offer a very compelling explanation for where all those "lost" jobs actually "went." As it has become increasingly obvious that neither Bagh nor the coward ever bother to read, let alone respond to anything published or posted that they don't like, preferring instead to just keep pounding on their scripted talking points, it's hard to justify paying them any more attention. And until they do begin actually to engage, I will not respond further. I suspect, therefore, I won't have much more to say on this topic to those two. Dafydd Posted by: Dafydd ab Hugh at March 12, 2004 06:07 PMYou go Dafydd! ::grin:: I had a link to this article in a comment here in another thread a few days ago and said about the same things you did. But not nearly as well. That's two threads now where this data has been dismissed or ignored. Posted by: Syl at March 13, 2004 07:41 AMPurveyors of the 'household survey is best' myth are once again forced to wear the conical hat and sit in the corner. Alan Greenspan (Forbes): 'Payroll Survey Is Best' WASHINGTON, Feb 25 (Reuters) - Federal Reserve Chairman Alan Greenspan repeated on Wednesday his belief the Labor Department's payroll survey was a better measure of job growth than the more optimistic household survey. "We have concluded that the data on so-called payrolls survey is surely the most accurate of the two and our suspicion is that at the end of the day there will be revisions to the household data," Greenspan said in response to a question from the House of Representatives Budget Committee. Of course, all those Marxists on Wall Street agree that the payroll survey is best--but they're probably anti-capitalist. Stephen Roach at Morgan Stanley: Nor do I buy the commonly expressed view that the data are simply wrong — that the payrolls-based survey of business establishments simply misses the inherent dynamism of risk-taking entrepreneurs whose enthusiasm for hiring can only be captured in the so-called household survey. Yes, the household-based job count is up 1.4 million workers in the 12 months ending January 2004 — well in excess of the paltry gain of 6,000 as measured by the establishment survey. According to the US Bureau of Labor Statistics, a little more than 25% of that discrepancy can be traced to definitional differences between the two surveys — namely a household survey that includes the self-employed, unpaid family workers, and private household staff. But the remainder of the difference could well be a perceptual one — disaffected workers sampled in the household survey who have downgraded their aspirations and simply won’t admit to the tougher reality depicted by businesses in the establishment survey. Yet there is really no comparison in the sampling accuracy of these two surveys. According to the US Bureau of Labor Statistics, the “active sample” of some 400,000 establishments in the payroll data covers about one-third of the total universe of such workers; by contrast, the monthly sample of only 60,000 households covers only 0.06% of the universe of over 106 million households in the United States. There is no doubt in my mind as to which of these two surveys should be trusted. Posted by: JadeGold at March 13, 2004 09:11 AMYet another pinko Maoist weighs in--this time the Chief Market Strategist for the Maxim Group , a global money management company. There are way too many head scratchers of the breed dismalus scientificus who are experiencing a false Eureka moment . . . they’re excited 'cause they think they found the dark matter, the missing link, the identity of Deep Throat. They think they found the “missing employed.” Even worse, the meme has been spreading. All manners of tortured data have been trotted out; The two different survey methods (business and home) have been challenged. Hey, we know they both suck (most surveys do), and the data is not exactly reliable, but at least there’s a baseline to compare it with historically . . . So now the latest brain droppings burst forth from both the Cato Institute and the AEI: Its the self-employed, work-at-home contractors! Thats the missing employed, and they don’t show up in any of our data! Eureka, all is well again! Ummm, in a word . . . No. About those so called "self-employed work-at-home contractors contractors:" THEY ARE JUST UNEMPLOYED.
Its the oldest dodge in the book. You adapt that phrase that allows you to power your way past the ackward pause that follows the “What are you doing at present?” question. There's different phraseology, depending upon your field: Attorneys claim they hung out their own shingle (“I have quite a few clients, but I’d like to get back to more challenging work than just closings and negligence litigation). On Wall Street, its called “Trading for your own account” (“Yeah, and I’ve been doing real well -- I can show you my P&L -- but I miss the cameraderie of working in a firm”) Sure, plenty of people legitimately work at home. But anytime I ever knew anyway who was "looking for work," they invariably described themselves as "self-employed/independent contractors./freelancers ." Apparently, the few remaining clueless who don't know what this euphemism really means are either jest plum ign'ant, or partisans trying to rationalize away the lack of job creation thus far in the recovery cycle. Posted by: JadeGold at March 13, 2004 09:53 AMJG...they just haven't been following the work of the BLS which is addressing all the 'problems' of the Household survey. You STILL haven't read any of the reports. they just haven't been following the work of the BLS which is addressing all the 'problems' of the Household survey. Baloney. Were it only politically-appointed economists--you might have an argument, albeit a vey weak one. The real problem you have is those economists and financial-types in the private sector. Their livlihood depends on the most accurate and timely info. To pretend they just don't understand or misinterpret the household survey is wishful thinking. [Folks] Until Coward #1 and Coward #2 coherently answer any of the points I raised in my last post, there is no reason for anyone to respond to their non-sequitur distractions. Just letting y'all know I'm waiting, half-smile on my lips, knowing what I will not be reading here any time soon. (Is anybody actually surprised that the two cowards haven't responded?) Dafydd First, Bagh's obsessive harping on the fact that the household survey is not corrected for population changes is complete nonsense, because the BLS has recently done exactly that, releasing new data series that do, in fact, smooth the month-by-month reports of the household survey for population shifts. It makes no difference; the gap between the two surveys persists, even when population changes are taken into account. Apperantly you did not read my post. I did not say that household survey is not corrected for population changes. My arguement is: The upward trend in household employment since the end of the 2001 recession has been largely a function of the estimated growth in population. That is to say, the household survey has not shown an increase in the proportion of the population that is employed. In fact, the employment-population ratio has declined since 2001: it was 64.3 percent at the start of the recession (March 2001) and 63.0 percent at the trough. The ratio declined further from late 2002 to 2003 and has been fairly flat over the past year. The increase in the household survey employment level, therefore, has been driven by the estimated growth in the population. Next, his fretting that the household survey is wanting because it surveys "only" 60,000 households (out of 100 million), while the payroll survey gathers data from 400,000 employers, is utter nonsense, as the most basic statistical understanding should have told him. I can only conclude, if Bagh is being honest, that he simply doesn't have even that cursory level of understanding of statistics. Go back and read my post again: Size of over-the-month change in employment required for a statistically significant movement: This is from BLS. If you admit that the payroll survey does not count contractors, new small business workers, and the self-employed, then can you estimate how many such workers there were in January 2001, how many there are today, and how many used to work for regular companies, hence were picked up the establishment survey... and who thus would falsely appear to have become "unemployed" according to that survey? It is true that establishment survey does not count the self-employed. But, those people do not "falsely become unemployed according to that survey" either, because the establishment survey gives you (surprise, surprise) the number of people on the payrolls. So? Easily resolved: the one that matches the other economic indices -- rapidly rising GDP, falling unemployment claims, falling unemployment rate, rising tax receipts, and business investment -- is almost certainly the one that is accurate. Wrong. Rising GDP, falling unemployment claims, falling unemployment rate, rising tax receipts, and business investment do not necessarily lead to a job growth. Rising GDP can be explained by productivity growth. Falling unemployment claims means less people are laid off (not an indication of hiring). Falling unemployment rate is due to decrease in the size of the work force. Business investment on capital goods may reduce the need for more labor. Second, the payroll survey has a history of massive revisions, often years after the original publication. Merely checking back a couple of months later, or even the year-end revision, doesn't even begin to fix the problems. The payroll survey has been revised recently, downward!!! "The authors of the Heritage Foundation report discovered that in fact, decreasing turnover alone accounts for between 1.0 million and 1.3 million job "losses" on the payroll survey." This is pure speculation. Anyone familiar with econometrics knows that their calculations are derrved from their assumptions on parameters and model specifications. That's the reason nobody takes this report seriously. Could you please answer the points I mentioned in previous posts: 1. Last month's participation rate of 65.9% was the lowest since September 1988. p.s. I don't appreciate being called "coward". Calling names is an indication of your intellectual level. [Bagh] Ah, at last an actual response! I retract the "coward" designation from you, for you are actually engaging -- in contrast to what you were doing before, posting numerous quotations that did not actually answer the call of the question. (But I will not retract it from JadeGold, who, you will note, continues to duck any and all tough questions.) > I did not say that household survey is not I had to read back a ways to figure out what you meant by this. By "function of the estimated growth in population," you mean that the Fed, trying to account for the unaccountable, has recently claimed that they may have overestimated illegal immigration, thus thinking there were more people than actually here; if they just naively multiplied the number who said they were employed by the ratio of surveyed to their total population estimate, and if the estimate were too high, then the employment number would likewise be too high. But the difference in the two surveys is three million. In order to get that huge a difference with about 66%-69% of the population working, they would have had to have overestimated the illegal-immigrant population by about four and a half million. As of 1997, however, the INS estimated that only about 275,000 immigrants illegally entered this country per year (http://uscis.gov/graphics/publicaffairs/newsrels/illegal.htm). Even if that rate dropped to ZERO after 9/11, that would only account for a discrepency of 687,500 in the population estimate between what the INS (hence the BLS) thought it was and what it really was... far too small to explain the jobs discrepency (if driven solely by overestimating population), which would require an overcount of 4.5 million. In fact, of course, you know as well as I that the rate of illegal immigration has not dropped to zero or anywhere near that. If there has been a slowdown -- and the border patrol doesn't seem to think there has been -- it has been very small, more on the order of a drop of 10%, not 100%. So we're talking about an overestimate of the illegal-immigrant population of less than 75,000. How, Bagh, do you propose that overestimating the US population by 75,000 will cause an overestimate of working Americans of three million? This explanation by the BLS is a canard. They have no idea what could explain the discrepency between the two surveys -- other than the obvious one, which they don't want to accept, since it would virtually invalidate one of their most treasured metrics. Population overestimates cannot explain the household survey showing three million more jobs than the payroll survey. >> Next, his fretting that the household So what? This is a non-sequitur: it is not logically connected to your point. If the difference between the household and the payroll surveys were 1% or 2%, you might have a point that the smaller sample size of the former might explain such a trivial discrepency. But this cannot possibly explain a wild difference like the former showing 800,000 jobs created since January 2001, while the latter shows 2.3 million jobs LOST. Nor can the difference in sample size explain why the discrepency is always in the same direction... we haven't seen a single instance during the recovery where the payroll survey showed more job growth than the household survey. Trust me on this mathematical point, Bagh: if one valid survey shows a net loss of 2.3 million jobs over the same time period another shows a net gain of more than three quarters of a million, then they are measuring different things. If we want to know which actually matches the reality of how many Americans are working and earning money, then we should check to see which jobs survey actually matches the rest of the economic data: and it is the household survey. It is the payroll survey that is wildly out of step with the other traditional measures of the economy. > It is true that establishment survey So if Joe Sixpack were employed by Ford Motor Company as a transmission designer at $35/hour six months ago, but today he is an independent contractor working at (but not for) Ford Motor Company as a transmission designer consultant at $48/hour, then the household survey will show Joe still employed. The payroll survey, however, will show that Joe left the payroll of Ford and didn't reappear anywhere... thus, he will look like he's unemployed when in fact he may be doing better than before. That is the problem with the payroll survey: if self-employment, independent contracting, limited-liability corporation creation (partnerships), and small-business employment is increasing -- as it has been markedly since 2000 -- then more and more people will disappear from payrolls, even though they're actually working and earning money. Thus the payroll survey will show huge job losses while the household survey shows the opposite... which is exactly what we're seeing now. Got it? >> Easily resolved: the one that matches No, but they're more consonant with job growth. It's very hard to imagine how all these positive economic indicators could continue for three years without creating any job growth at all. Even so, if both surveys were showing huge job losses, you would have a good point. But they don't; we have two contradictory survey results, and they can't both be accurately measuring the same thing. The whole point of citing such alternate economic criteria is not to claim this as proof that job growth is occurring, but rather to decide which survey result is more believable. If other indicators were there for an economic recession, then I would say the indicator that showed job losses was the more believable. But in the current case, all the other indicators show a robust, growing economy... so the traditional indicator showing job growth is inherently more believable than the traditional indicator showing job losses. That is, the burden of proof now shifts to the defenders of the payroll survey to explain why their survey shows so many fewer jobs than the household survey, not the other way around -- as it would be if there were a recession and the household survey showed huge job growth (which it never has, by the way; during the actual recession, the household survey as well showed job losses; the pickup occurred during the recovery, not the recession). > The payroll survey has been revised And upward, and sideward. That is my point: it is a very volatile metric that is adjusted all over the place even years after the fact. The payroll survey is not as good a measure as it used to be and no longer as good as the household survey in recent years. > Could you please answer the points I Of course I will. > 1. Last month's participation rate of According to...? Without a citation, preferably with a link, I have no idea what table you're looking at. I will be happy to respond when you give me a link. > 2. The average unemployment spell is Every time you increase the length of time that unemployed people receive unemployment, you will likely increase the "average" (mean? median?) length of time that people remain unemployed, because there are always people happy to be paid for doing nothing as long as they can. My friend Jim, for example, who admitted to me that he never seriously tried to look for work until his unemployment compensation ran out. (And note also that at some point, people "fall off the back end" of any measure of unemployment that measures only those "actively seeking employment." If we used to have a backdoor that cut people off earlier -- as example, when we paid compensation for a shorter period of time -- then that would artificially reduce the "average" time people spent officially listed as unemployed. If now we count them longer, that would automatically increase the "average" time they spent being officially unemployed. Some portion of this "rise" is purely an artifact of the way the survey is conducted. I also want to know the source of this number: was it calculated from the payroll survey data? The household survey data? Direct unemployment-compensation data? A separate survey? Also, without seeing the number in context, I have no idea whether this is a trivial change or a serious change. Once again, can you please source your data? > 3. The 1.6% annual growth rate of the Assuming your figures were not already adjusted for inflation, all you are showing is that the growth that is occurring is not in blue-collar, non-managerial manufacturing jobs. This is hardly a startling observation. Our economy is phasing out manufacturing, as are the economies of every advanced nation on the planet and even some developing nations (Red China, for instance), in favor of computerized automation... and there is nothing any president could do to change this, short of outlawing computers. This is not an original observation [g]. The question is whether the workers who used to have a job sticking labels on beer bottles as they went by on an assembly line, who have recently been replaced by a beer-label-sticking machine, have found employment elsewhere for comparable wages -- as the household survey suggests -- or are starving to death in lonely shacks outside Milwaukee, as the payroll survey suggests. Bagh, you and are in in full, 100% agreement that traditional, blue-collar manufacturing jobs are disappearing. Where we disagree is that you see a horrific situation that needs John Kerry to rectify it (bring back those dark, satanic mills!), while I see it as the inevitable consequence of a shifting economy... just as the percent of American workers doing stoop labor in the fields has plummeted since 1850. This is a fact of modernity: things change, fluctuate, and workers in dying industries have to scramble to find somewhere else to go. You cannot put a time-bubble around America and stop the decline in manufacturing... any more than you could have put a time-bubble around Detroit and kept out all them ferrin cars. When you try to do that, you get -- well, the Soviet Union, North Korea, and Cuba. I would rather live here in the vibrant United States, thankyouverymuch. Dafydd ab Hugh Posted by: Dafydd ab Hugh at March 16, 2004 06:11 PM[Bagh] Finally, Bagh, you have signally failed to address the most original contribution of the Heritage report: the discovery of a massive underestimate of employment on the payroll survey due to changing patterns of job turnover. I wrote: > The authors of the Heritage Foundation Your only response was to say, yeah, well, that's just their estimate. Fine. What's wrong with their estimate? What's your estimate of the number of jobs falsely "lost" due to reduced turnover? Do you argue that the phenomenon doesn't happen at all? Why not -- the Heritage report went into great detail on exactly how they calculated the number they did... what specifically do you object to in their calculations, and what is your standing to object to it? At least Heritage was quite up front about where they derived that number. You have done nothing to respond to this point except wave your hands. Read the report -- which I suspect you still have not done -- and tell me where, exactly, they went wrong. I will not accept vague generalities: if you think an assumption is wrong, cite it specifically and tell us why you think it's wrong. Then give a better or corrected version and recalculate the number. (The math isn't that hard; nothing beyond high-school algebra required.) Well, Bagh? Dafydd Posted by: Dafydd ab Hugh at March 16, 2004 06:25 PMBy "function of the estimated growth in population," you mean that the Fed, trying to account for the unaccountable, has recently claimed that they may have overestimated illegal immigration... So we're talking about an overestimate of the illegal-immigrant population of less than 75,000. I agree that overestimating illegal immigration alone is not sufficient to explain the difference between payroll and household survey. However, according to BLS the difference might be in the amount of hundreds of thousands, if not millions!: "As part of its annual review of intercensal population estimates, the U.S. Census Bureau determined that a downward adjustment should be made to the household survey population controls. This adjustment stemmed from revised estimates of net international migration for 2000 through 2003... A comparison of December 2003 data based on the old and new controls indicated that the population decrease caused declines in the labor force (-437,000), employment (-409,000), and unemployment (-27,000)." " This latest population control adjustment follows two previous increases in the controls. With the release of January 2003 household data last year, BLS introduced two separate adjustments that increased the survey population controls. Both of these adjustments in population controls resulted in an increase in the employment estimates from the household survey. The impact on employment of the January 2000 adjustment was about 1.6 million. The impact of the January 2003 adjustment was about 576,000. Source: http://www.bls.gov/cps/ces_cps_trends.pdf This explanation by the BLS is a canard. They have no idea what could explain the discrepancy between the two surveys -- other than the obvious one, which they don't want to accept, since it would virtually invalidate one of their most treasured metrics. I don’t understand why the BLS and the Federal Reserve Bank would engage in distorting the situation in the labor market. If the difference between the household and the payroll surveys were 1% or 2%, you might have a point that the smaller sample size of the former might explain such a trivial discrepancy. The difference between the payroll and the household surveys is in fact as little as 2 %. From March 2001 (peak)- February 2004: Difference is 2 562 000. Total employed population approx. 130 million. If you recall, in your previous post you accused me of being either dishonest (?) or “not having even that cursory level of understanding statistics”. I showed that the sample size makes a difference in terms of the precision of estimates. You also mistakenly compare employment surveys with “polls of a mere 1 500” which are used “to reflect a voting population of 180 million with a margin error of only 4%”. Those polls are not used to estimate the number of votes of candidates. They are used to estimate the percent of population voting for a candidate. 4% margin of error means more than 4 million votes. As I mentioned above we are talking about a 2 percent difference between payroll and the household surveys. So if Joe Sixpack were employed by Ford Motor Company as a transmission designer at $35/hour six months ago, but today he is an independent contractor working at (but not for) Ford Motor Company as a transmission designer consultant at $48/hour, then the household survey will show Joe still employed. The payroll survey, however, will show that Joe left the payroll of Ford and didn't reappear anywhere... thus, he will look like he's unemployed when in fact he may be doing better than before. That’s true, because it is not payroll survey’s purpose to measure the number of self-employed population. Could you also please give me a link that supports your argument, that is, those self-employed are better off being self-employed? It's very hard to imagine how all these positive economic indicators could continue for three years without creating any job growth at all. Three years? Are you sure? GDP growth in the last three years was 0.5, 2.2, 3.1%. Average unemployment claims have been below 400,000 only since 2003Q4. 2001 Q4: 446,385 (SA) The economy was not growing significantly" untill the third quarter of 2003. Only in the last two quarters it shows a large growth. And upward, and sideward. That is my point: it is a very volatile metric that is adjusted all over the place even years after the fact. The payroll survey is not as good a measure as it used to be and no longer as good as the household survey in recent years. Both surveys are adjusted regularly. Adjustments in the household survey are usually in the range of hundreds of thousands “The impact on employment of the January 2000 adjustment was about 1.6 million. The impact of the January 2003 adjustment was about 576,000”. You did not answer the following point: New business births in the payroll survey – The payroll survey sample does not include new firms immediately. They are incorporated with a lag. In the interim, a model-based estimate is used each month to account for employment resulting from new firm births. Based on the relatively small benchmark revision for March 2003, as well as comparisons to universe data for Second Quarter 2003, the model appears to be performing well during the recent period.” http://www.bls.gov/cps/ces_cps_trends.pdf > 1. Last month's participation rate of Oh please. Do I have to go to BLS side on your behalf and download statistics? http://data.bls.gov/servlet/SurveyOutputServlet?data_tool=latest_numbers&series_id=LNS11300000 2. The average unemployment spell is Every time you increase the length of time that unemployed people receive unemployment, you will likely increase the "average" (mean? median?) length of time that people remain unemployed, because there are always people happy to be paid for doing nothing as long as they can. Any study supporting your point? Any link? I also want to know the source of this number: was it calculated from the payroll survey data? The household survey data? Direct unemployment-compensation data? A separate survey? I am shocked, shocked! Don't you know that BLS has a beautiful website. It is the household survey. Go to http://www.bls.gov/webapps/legacy/cpsatab9.htm. > 3. The 1.6% annual growth rate of the Assuming your figures were not already adjusted for inflation, all you are showing is that the growth that is occurring is not in blue-collar, non-managerial manufacturing jobs. It is similar in both goods-producing private sector and service-producing private sector. Please, please go to http://www.bls.gov/webapps/legacy/cesbtab4.htm. Finally, Bagh, you have signally failed to address the most original contribution of the Heritage report: the discovery of a massive underestimate of employment on the payroll survey due to changing patterns of job turnover. I wrote: The parameters they used: Let’s start with turnover decline: If you look at the footnote 27 they stated that After reviewing an early draft of this paper, economists at the Bureau of Labor Statistics suggested that the BED may not be appropriate for gross worker-flow turnover. The alternative suggested is an employer-to-employer worker flow from CPS surveys. These data show a decline from 2.89 percent per month in November 1999 to 2.09 percent in November 2003— half of the rate derived from the BED, and still large. The CPS number is not definitive, however, and potentially underestimates the effect by ignoring respondents who drop out of the sample. This attrition bias is likely to be made up of a large proportion of individuals who move in response to employment change. What happens if we used CPSD instead of BED? Then turnover decline would be 0.8%. Put it into equation and payroll deflation would reduce from 1 768 000 to 832 000. Next, let’s look at Re-employment rate. I quoted footnote 30 below: Using very complicated CPS calculations, Fallick and Fleischman estimate a comparable re-employment rate of 40 percent. This estimate is a lower bound, and the authors discuss an “attrition bias” that would result in an upper bound rate of 80 percent. Of course, other estimates based on other methodologies will offer different rates. All researchers in this area acknowledge that there is no definitive study of worker flows, and the one clear fact is that employer-to-employer transitions are much more prevalent than economists realized a generation ago. If we use Fallick and Fleischman’s estimate in the base model, which is 40 percent, the payroll deflation would be 884 000. Combine this with the lower estimate for turnover decline (0.8 percent), the payroll deflation becomes 416,000. That’s to say payroll deflation is something between 416 000 and 1 768 000. One more point. The authors calculated the payroll deflation since 1999 . According to the BNA, Inc., which they referred on page 12 to emphasize the importance of turnover rate, that rate was 1.3 percent in 2000, 1.1 percent in 2001 and 1.0 percent in 2002. Assuming that turnover rate further declined to 0.9 in 2003, the decrease is just 0.4 points, not 1.7. Satisfied? I am not blaming Bush either for the recession (the economy slowed down in the third quarter of 2000 and fell in recession in the first half of 2001), nor the jobless recovery. It is quite clear the loss of jobs is structural - nothing to do with the Bush's economic policies. Americans should also credit his tax cuts and FedResBank's monetray policies for this mild recession and quick recovery (I should also point out that his fiscal policies will cause many problems in the next few years. I believe the tax cuts should have been temporary, more oriented to middle class, and coupled with incentives for job creation intstead of encouraging capital investment). The Republicans should emphasize their successes instead of trying to defend the indefensible, i.e. the economy has created jobs. It has not - yet. Most of the job losses are not their fault, but they are going to get blamed for it anyway. This is politics. A famous economist once said if you torture the data hard enough, it will confess anything. This is what you and some other people in this blog are doing. All economic indicators point out a sluggish labor market (labor participation rate, wages, duration of unemployment, capacity utilization rate). They payroll data is consistent with that and this is why everybody is using it. p.s. I have another link for you for wage growth. http://www.epinet.org/content.cfm/webfeatures_snapshots_archive_10292003 "Real wage and salary income declined by 1.2% between the start of the last recession (March 2001) and the most recent month of data (August 2003). That decline is the worst performance at this stage since the current data series began in 1959. " p.s. I have to finish my Ph.D thesis - that's why I post infrequently. Posted by: bagh at March 17, 2004 01:26 PMI also forgot this one. This is also from BNA cited in that Heritage Foundation Report: http://www.wagetrendindicator.com/highlights.htm Wage Index Falls to Lowest Level Since 1993 An even weaker labor market than observed a month ago has resulted in the lowest reading of BNA's Wage Trend Indicator™ since early 1993, according to the latest report released today by BNA, Inc., a Washington, D.C. -based publishing company. The latest decline in BNA's Wage Trend Indicator™ signals that private employers can expect to pay annual wage increases of close to the prevailing 3 percent level through the rest of this year, according to economist Joel Popkin, who developed the WTI for BNA along with economist Kathryn Kobe. Just as it was in 1993, the U.S. economy is stuck in a period of tepid job growth that has dampened wage pressures in private industry, as the forward-looking WTI illustrates. “The labor market is about as weak now as it was in 1993, and the rate of annual wage increases is about the same as it was 11 years ago,” Popkin said. The revised reading of BNA's Wage Trend Indicator™ pertaining to the first quarter of 2004 is 98.38, the lowest since the first quarter of 1993 when it stood at 98.35 (second quarter 1976=100). As released last month, the preliminary reading was 98.51. It is me again. It would be a great injustice to blame Bush's fiscal policies (for the fiscal disaster that will occur in a few years) while ignoring Greenspan's monetary policies which have caused enormous asset bubble and incredible levels of debt in the household sector that would bring further disaster to the U.S. economy when the interest rates go up. When historians look back this era years later they will conclude that this was the most expensive recovery that money can buy. p.s. Warning: The U.S. economy is known with its resilience and raising from the ashes even stronger. For that reason the predictions above, which would be valid for any other country in the world under similar circumstances, may not happen. Posted by: bagh at March 17, 2004 02:07 PM[Bagh] >> If the difference between the You know, as I wrote that, I thought to myself, I wonder if Bagh is going to try the old trick of trying to minimize the change by comparing it to the entire population, rather than to the change? Sure enough! But heck, why stop with the US population? As long as we're comparing to irrelevant population groups, why not look at both surveys as a percent of the world population? That yields a difference of a mere 0.1%... why, the two surveys are identical! The percent I'm talking about is the difference between the two surveys with respect to each other, Bagh: the payroll survey shows a drop of about 2.5 million from the beginning of 2001. The (unadjusted for population revisions) household survey shows a rise of about 800,000 over that same period... that is a difference of 132% of the change -- all of the drop plus an additional rise... and consistently over time with an accelerating rate. That is the discrepency we're talking about. > Source: http://www.bls.gov/cps/ces_cps_trends.pdf Aha, thank you for providing that link. Now let's take a look at that PDF file. Take a look at page 5, the chart comparing the two surveys side by side... and notice what the legend says about the household survey: "the household series [green line] presented here has been smoothed for population control revisions. The 'adjusted' household series [red line] has been smoothed for population control revisions and adjusted to an employment concept more similar to the payroll survey." (The blue line is the payroll survey.) (Emphasis added.) Now, notice the trends of the three lines: The population-revision-adjusted household survey (green line) -- that's after taking into account the population adjustments that so exercised you in the first part of your first post -- hits a post-recession nadir in January 2002 at about 135.8 million jobs. From there until February 2004, that survey shows a rise of about 2.4 million to 138.2 million. That's after adjusting out the population revisions. The household survey that is also adjusted to change the definition of "employed" to more closely match that of the payroll survey (red line) also hits a nadir on January 2002 at about 128.8 million; from there until February 2004, it rises to about 130.4 million, an increase of 1.6 million. Finally, the payroll survey (blue line) on January 2004 was at approximately 130.5 million; but on February 2004, it was about 130.0 million, a drop of about 500,000. Thus we analyze: the original survey shows a rise of about three million from the nadir. When you adjust for population revisions, that turns into a rise of about two and a half million. So approximately 500,000 of those jobs created (on the unadjusted household survey) are due solely to revised population estimates... BLS overestimated the population, hence overestimated job growth on the household survey: the rise should have been 2.5 million, not three million. But next, when you switch employment definitions to more closely match what the payroll survey measures -- that is, excluding agriculture, self-employment, contracting, and LLCs -- you still show a rise in employment from the nadir... but the rise is only one and a half million. So that means a discrepency of about one million jobs is due solely to the rise in types of employment that the payroll survey simply doesn't capture. Finally, even that survey differs from the payroll survey; the fully adjusted household survey -- population revisions and also type of employment -- shows a rise of one and a half million; but the payroll survey shows a decline in employment during this same period of about half a million. (It's hard to compare using the actual numbers at the end of the document because they don't cover the same periods; but using the much shorter period of April 2002 to December 2003 -- the only period covered by the numbers given for all three survey metrics -- we have the population-adjusted household survey showing a net rise of 1.8 million jobs; the population and employment-definition-adjusted household survey showing a net rise of 0.6 million; and the revised payroll survey showing a net loss of 0.4 million. So just during this abbreviated period, types of employment not covered by the payroll survey rose by 1.2 million jobs, and there is a discrepency in the total job picture between the household and payroll surveys, even adjusted for population revisions, of 2.2 million jobs.) Here is the important point: the rise in population cannot explain the discrepency between the two surveys, because both purport to show total employment: the household survey shows that it has been keeping up with population growth... but the payroll survey shows that it has not been, that it has declined significantly even as population has grown significantly. And even when the BLS makes its best attempt to revise population estimates, at the end of the day, the household survey shows a net increase in total working Americans during this administration, while the payroll survey shows a large net decrease. Increasing population cannot "account for" this discrepency; both claim to show total jobs. That difference cannot be explained either by population increase or by the fact that the payroll survey doesn't measure some types of employment; it is caused by something that is an artifact of the payroll survey itself. And the only candidate put forward so far to explain this difference is the false decline caused by the decrease in job turnover: turnover results in overcounting of jobs on the payroll survey; so as turnover declines following a recession, the amount of overcounting declines, so it falsely appears as though the number of jobs has declined. This data dovetails very nicely with the independent analysis done by the Heritage Foundation. Nice to see the BLS forced to admit the problems with the payroll survey are not merely skin deep. There is also this tidbit from page 9 of your source: "As was illustrated in Box 3 above, adjusting for the measurable differences in the surveys' employment definitions resolves only a portion of the discrepency. This adjustment process is imperfect, however, because precise data are not available in many cases to make the best possible adjustment. For example, some independent contractors are not reported as self-employed in the household survey, but rather as wage and salary workers. This type of reporting issue limits BLS' ability to fully reconcile the two employment measures." Yes, BLS is correct: the problem of not counting the self-employed only explains a part of the discrepency... but that part is a minimum of 500,000 (20%), and as they coyly noted above, likely a lot more -- since studies (cited by Heritage) show that about as many independent contractors report themselves as "company employed" as the number of people who report themselves as "self employed" altogether... which means that the inability of the payroll survey to count those not on a payroll in fact accounts for, by itself, a rise of one million jobs that is not seen by the payroll survey. (The BLS also notes that workers who are paid "off the books" will show up on the household survey but not the payroll survey as well; thus, if we simply lump this problem into the general problem of the payroll survey not picking up many of the employed, that increases the "payroll undercount" beyond a million jobs. But how far beyond is anybody's guess.) > I don’t understand why the BLS and the Then you are the only one, Bagh, who doesn't understand that huge bureaucracies will scramble to find any explanation for a glaring error other than the one that would cause them to completely retool their entire method of doing their jobs. Look how long the Pentagon has continued the obviously flawed procurement system, despite study after study showing that it doesn't work well. That is a hallmark of bureaucracies. > Could you also please give me a link Better off than being unemployed? You actually need a LINK to tell you that it's better to be earning money than not to be earning money? Yeesh! To be charitable, I will assume that what you meant to ask for was data indicating that it's better to be self-employed than employed by a company. But I never made that argument, because it completely depends upon the individual circumstance. > Oh please. Do I have to go to BLS side No, Ace; you need to present enough information to put the data you cite into context. The drop in participation rate during the recovery is about 0.5% (from 66.4% in January 2002 to 65.9% in February 2004). But where is this number coming from? Clearly this comes from the payroll survey, not the household survey. How do we know? Very simple: the figures you cited above show a participation rate in March 2001 of 67.1%; but the PDF you cited earlier, quoting from the household survey, says "it was 64.3% at the start of the recession (March 2001)...." So clearly, these new figures you cite are from the payroll survey, not the household survey. We have no data about what the household survey shows, other than that the participation rate has been "fairly flat over the past year." But the figures you cite were not "fairly flat" over the last year; in fact, the participation rate shows a consistent decline during this period. So when you say "last month's participation rate of 65.9% was the lowest since September 1988," you're talking about the payroll participation rate. On this we agree: payroll employment is decreasing as a percent of the population. But your conclusion is that there are millions of huddled, starving masses, while the Heritage Foundation's conclusion -- based upon actual data -- is that there are millions of people working in jobs that the payroll survey doesn't see... and another million fake jobs that never existed except as a payroll-survey overcount. > 2. The average unemployment spell is How odd... when I go to the Website you cite -- http://www.bls.gov/webapps/legacy/cpsatab9.htm -- and ask for the "average" (they mean the mean) weeks of unemployment, I indeed get 20.3 weeks. But when I ask for the median length of unemployment -- the point where half the people are unemployed longer while half are unemployed shorter -- I get the number 10.3 weeks... which is considerably lower than it was in June of 2003, 11.7 weeks! The mean is a problematical number because it's strongly affected by numbers with a high standard deviation. For example, if you have a hundred people in the room, 99 of whom make $40,000 per year, but the other one is Bill Gates, then your median income for that group is $40,000 per year, but your mean income is about $30 million per year. Which figure more accurately represents that group? Now, let's look at the legislative side of the picture. Here is an interesting point to ponder: "On January 8, 2003, the U.S. Congress approved President Bush's request for a renewal of the Temporary Extended Unemployment Compensation Program (TEUC), which extends the normal 26-week limit on receipt of UI benefits. This renewal continues funding for the TEUC that began in March 2002 and expired in December 2002." http://www.frbsf.org/publications/economics/letter/2003/el2003-05.html In other words, starting in March 2002, the feds increased the duration of unemployment compensation from 26 weeks to 39 weeks in all states and 52 weeks in "high unemployment" states. Now let's compare to the duration of unemployment. From the beginning of the recession, March 2001, to March of 2002 (just before the expansion went into effect), the mean time spent on unemployment rose from 12.9 weeks to 15.5, a rise of 2.6 weeks; this makes sense, because the household survey (adjusted for population revisions), the source for this unemployment figure, shows a decline of 1.5 million jobs (that's the recession). But from March 2002 until March 2003, the mean unemployment duration also rose by an identical amount, 2.6, to 18.1... despite the fact that on the population-adjusted household survey, there was an increase of 663,000 jobs! In other words, at the very time that jobs were expanding, the length of time unemployed people spent on unemployment compensation rose by 2.6 weeks. From March 2003 to February 2004 (the most recent figure), mean duration rose by an additional 2.2 weeks... so after enacting an extension of unemployment benefits, the mean length of time on unemployment increased by 31%. The increase is not as dramatic when you look at the median (indicating that the results are, in fact, being skewed by a small number of people staying a very long time on unemployment compensation): From March 2001 to March 2002, the median duration increased 1.8 weeks; from the extention to March 2003 it increased by 1.3 weeks; and from then until now it increased by only 0.6 weeks, for a total increase after the extension of 23%. Easily explained mathematically: from the extension in March 2002 until now, most people's time on unemployment didn't increase much, if at all; but a smallish number of people's duration increased markedly. This leads to a total increase in the median of 1.9 weeks, but an increase in the mean of 4.8 weeks, two and a half times as much. That's a top-heavy graph, like the room above full of average Joes, plus one Bill Gates. There are several possible explanations for this, among which are people taking advantage of the extension of unemployment benefits, and also perhaps a hard-core of un-retrainables growing increasingly disengaged from the workforce... which is pretty much what you expect as the economy changes. After a while, more of those people retire to be replaced by younger, more flexible workers, and employment increases. This happens again and again as the workforce changes; think of the 1970s in the automobile industry. >>> 3. The 1.6% annual growth rate of the Now that's odd... when I go to the Website you linked above and select for manufacturing, I get a rise of 2.4% in mean wages for the last year, more than twice the inflation rate of 1.1% (by the CPI). When I pick all goods-producing work, I get a rise of 2.3%. When I pick all private service-providing work, I get a rise of 2.0%, both about twice the inflation rate. And this is from the establishment survey! It doesn't count the self-employed, contractors, people working for startups, and people in LLCs. How did you get a figure of 1.6%? (And even that weird figure is higher than the inflation rate.) Workers are not "falling behind in real terms," as you claimed. > (Lengthy quotations from Heritage ...With the mean at 1.1 million -- exactly what Heritage estimated. What's your point? That you agree with the Heritage Foundation that the payroll survey overstates job losses by around a million? All right, I accept your agreement. If you add that million to the million people that even the BLS admits are earning income but aren't counted by the payroll survey, then add in the equivalent number that the studies cited by the Heritage report indicate are also working but not counted by the payroll survey (BLS admits this class exists but declines to estimate its size), you get three million... just about the discrepency between the surveys. Haven't you just admitted my whole point? > One more point. The authors calculated This, Bagh, is an example of "cherry picking." You scanned and scanned and scanned and finally found one source that was "referred" to in the report -- referred as in Heritage mentioned that somebody else mentioned BNA -- which estimates a lower turnover rate... so you promptly seized upon this as a repudiation of some sort of the entire paper. (You found this by Googling on the Bureau of National Affairs, going to their Website, then searching on "turnover" in the text box and clicking on the first hit you got.) But this puts you in the position of comparing rates from different sources, which may or may not use the same definitions, and which you found by noodling on the Web for a few minutes, for the sole purpose of smashing down the knowledgeable conclusions of legitimate researchers in carefully conducted studies. This is just silly, Bagh. Please explain the discrepency between the calculation of turnover rate by the Business Employment Dynamics (BED) data series by the BLS, the Current Population Survey (CPS) by the BLS, and the estimate by the Bureau of National Affairs, Inc. (BNA), and why you think the BNA is the best estimate -- apart from the fact that it yields the result you prefer. I appreciate you taking the time finally to read the Heritage Foundation report, Bagh. But you are still in the mode of trying to cherry-pick some little point here or some datum there in order to "prove" that the payroll survey hasn't been undercounting job creation during this recovery, rather than honestly facing the fact that there is a very strong liklihood that it has. Do you not agree that it's important to discover whether the payroll survey has become a "bad source" recently? Do you not agree that the important thing is not to defend this or that survey, but rather to determine the actual reality of what is happening in the job market? If so, then instead of digging in your heels to swat down the Heritage Foundation report, why don't you contact the authors directly with your concerns and read their reasons for drawing the conclusions they drew? The Heritage Foundation, while leaning slightly to the right, has no history of producing false or bogus reports for partisan purposes; in fact, their reputation among economists on both sides of the aisle is pretty high... as indicated by the willingness of, e.g., the Bureau of Labor Statistics to formally review a preliminary version of this article. Waving them away as if they were nothing (and offering nothing in exchange except for what you can grab from the Internet without too much time or effort) is not simply wrongheaded, it borders on the irrational. I don't know if you have a degree in economics; I don't. But my degrees in mathematics do teach me that when recognized experts in a field raise a significant point that could change our whole picture of some phenomenon, the proper response is to take it seriously and see what other experts in the field (who have read the previous experts' paper) have to say about it... not to find some reason to completely dismiss it because it doesn't comport with your pet political yearnings. Dafydd ab Hugh Posted by: Dafydd ab Hugh at March 17, 2004 05:43 PM[Bagh] And by the way, I am not claiming that you are in the camp that blames Bush for everything. But you clearly are in the camp that believes in soaking the rich (you say the cuts should have been "temporary" and targeted at the "middle class," which you don't define). You are, in fiscal terms, a moderate liberal. Say about the level of John Kerry, no surprise. You don't believe that cutting the taxes of the rich (however you define that) helps grow the economy. Well I do, and I believe the extremely strong growth in the last few quarters clearly shows that. You believe the payroll survey because it shows a stagnant job market. You believe BNA because they predict a stagnant job market. It may only be coincidental that this "stagnant" job market would hurt Bush; but it certainly hasn't escaped your attention, as you note that Bush will be "blamed," rightly or wrongly. As such, you are as partisan about this as anyone here on the board; so when you talk about "torturing" the economic data, you should be honest and admit that that is what you are doing as well. But there is a real reality out there; and in the real reality, real people are reporting that they're working and jobs are growing, even if your favorite survey doesn't show it. And the odds are that even the lagging payroll survey will start to move fairly soon. When it does, will you be back here trying to explain why it's no longer a good measure, and we shouldn't believe in job growth even if it comes from the payroll survey? Or will you admit that jobs were being created all along, and it just took corporate payrolls a while to catch up? Dafydd You know, as I wrote that, I thought to myself, I wonder if Bagh is going to try the old trick of trying to minimize the change by comparing it to the entire population, rather than to the change? It is getting difficult to follow you. I did not compare the difference with the US population or world population. I compared it with the number of people employed. Let X denote this number. For the sake of argument let’s assume that HH survey correctly measures the number of people employed, whereas the payroll data measures it with an error (let’s call it “e”). If you want to measure the magnitude of this error you divide “e” by X. This is how it works in statistics. Your method is flawed. You are dividing the difference between the surveys with each other (I guess you divide 3.3 million – the difference - by 2.5 million – the increase in HH survey). The problem is if the difference were just 3 and the HH survey showed an increase of 2, you would get almost the same result – a difference of 150%. Here is the important point: the rise in population cannot explain the discrepancy between the two surveys, because both purport to show total employment…And even when the BLS makes its best attempt to revise population estimates, at the end of the day, the household survey shows a net increase in total working Americans during this administration, while the payroll survey shows a large net decrease….Increasing population cannot "account for" this discrepancy; both claim to show total jobs….That difference cannot be explained either by population increase or by the fact that the payroll survey doesn't measure some types of employment; it is caused by something that is an artifact of the payroll survey itself. You clearly did not understand the rationale behind focusing on population controls to explain the problem with HH survey. Let me try it once more. Nobody is saying that household survey is not corrected for population changes. In that BLS report they mentioned that “This version of household survey employment smoothes out the effects of sizeable population control revisions to the survey in January 2003 and January 2004.”. So, they already revised HH employment figures to make them comparable across time. That’s not the issue here. Let’s assume that the population consists of three groups of people. Employed (X), unemployed (Z) and not-in-the-labor force (Z). The payroll survey focuses on the employed, X. More specifically, a subsample of X – the number of people in the payroll of non-farm establishments (Let’s call this group X1). The sample covers “160,000 businesses and government agencies, which cover approximately 400,000 individual worksites drawn from a sampling frame of over 8 million Unemployment Insurance tax accounts and nearly 97 percent of total nonfarm employment.” That’s why professionals have a high regard for the payroll survey. It covers 97 percent of X1. http://www.bls.gov/ces/cesmetho.htm. HH survey, on the other hand, collects data from a sample of approximately 60,000 households. The sample includes people in X, Y and Z. Let x, y, and z denote the number of people covered in the HH survey who are employed, unemployed, not-in-the-labor-force, respectively. Next, using this sample the BLS calculates the ratio of population who are employed, unemployed, not-in-the-labor-force. For example, y/(x+y) gives you the unemployment rate, y/(x+y+z) gives you labor participation rate, etc. Since 60,000 is a pretty large sample, the margins of error in these ratios are very small. For example, a 0.2 percentage point in the unemployment rate is significant at a 90-percent confidence level. http://www.bls.gov/opub/hom/homch1_f.htm. When it comes to making inference about the whole population, you have to multiply x,y, and z with some population control. For example, let’s say 30,000 people are reported to be employed in the HH survey. You multiply this number (x) with some population control (let’s call it k) to estimate X. The population controls are developed by the U.S. Census Bureau. They are derived from decennial census information and, between census years, from administrative and other data. There are limitations to the population control estimates due primarily to the difficulties associated with estimating the net international migration component. The population controls contributed significantly to the discrepancy between payroll and household survey employment in the 1980s and 1990s when the household survey showed less growth than the payroll survey. About the use of data: Payroll survey produces nonfarm payroll estimates for: all employees, women workers, production workers, average weekly hours, average hourly earnings (constant dollar and current dollar), average weekly earnings, average overtime, index of aggregate hours and payrolls, and diffusion indexes. HH survey is the source of statistics on the activities of the labor force, including unemployment and the Nation's unemployment rate. http://www.bls.gov/ces/cescope.htm. Pros and cons of the two surveys: Payroll data, as I mentioned above, focuses on a subsample of the employed population. It does not cover self-employed, work force in the agriculture etc. However, the estimates on non-farm payroll numbers are very reliable. HH survey looks at the overall labor market and gives you a picture of ratio of employed and unemployed people in the population as well as the ratio of people in the labor force. Measuring the the level of employment is not its purpose. Why? Because of small changes in the population control (k) may cause big variations. That’s one of the reason of large margin of error inherent in this survey when you try to make inference about the level of employment: Size of over-the-month change in employment required for a statistically significant movement: Both surveys, from time to time, are revised. The adjustment in the HH surveys comes from adjustments to underlying population base, which s revised annually to intercensal estimates, and every 10 years to the decennial census.. For example, in January 2000 adjustment the employment was revised upwards by 1.6 million, in Jan 2003 upwards by by 576,000, in Dec 2003 downwards by 409,000. The problem with the HH survey is that it is benchmarked only once a decade to the decennial census, resulting in a less precise employment measurement than the payroll survey. The payroll survey is adjusted due to change in company numbers. It is benchmarked annually to employment counts derived primarily from Unemployment Insurance (UI) tax records. At the total nonfarm level, the March 2003 benchmark revision was a downward adjustment of 122,000 or -0.1 percent. According to the BLS: Based on the relatively small benchmark revision for March 2003, as well as comparisons to universe data for Second Quarter 2003, the model appears to be performing well during the recent period. As a summary,
This is a purely a speculation. If HH survey and payroll survey were conducted separately by two different agencies, you might have a point. It is not the case. If BLS put more weight one survey (in this case payroll) in the expense of the other (HH), then it is admitting to have problems with the HH survey and undermining its own status (using your words, it has to “completely retool their entire method”). Even if I give your argument the benefit of doubt, then what makes other federal agencies (including the commerce and Federal Reserve Bank) support the same thesis? Just recently, Greenspan in his testimony to the Congress made the following statement: "I wish I could say the household survey were the more accurate. Everything we've looked at suggests that the data on so-called payrolls survey is surely the most accurate of the two and our suspicion is that at the end of the day there will be revisions to the household data." "To test the self-employment theory, the Fed adjusted the household survey by taking out all the kinds of workers who do not show up on the payroll survey - self-employed people, but also farm workers and family workers in family-run companies. Even then, Mr. Greenspan said, the discrepancy remains large. The Fed's conclusion was that the household survey's results have been inflated by overestimates of population growth. Because the household survey is a sample, the Bureau of Labor Statistics infers the total change in jobs by multiplying the ratio of employed to unemployed workers in the household survey by its estimate of the total population. If the population estimate is too high, the estimated number of jobs will also be too high. > Could you also please give me a link Better off than being unemployed? You actually need a LINK to tell you that it's better to be earning money than not to be earning money? Despite my poor choice of words, I am sure you understand what I meant. In your previous post, you made the following statement: So if Joe Sixpack were employed by Ford Motor Company as a transmission designer at $35/hour six months ago, but today he is an independent contractor working at (but not for) Ford Motor Company as a transmission designer consultant at $48/hour, then the household survey will show Joe still employed I read your statement as an argument in favor of self-employment or contract working as opposed to being employed by a company. Since that was not your purpose, I take it back. The reason I raised the issue is that even we put the “level of employment” aside, the quality of jobs has been deteriorated in recent years. ”Over the past two years, expanding industries paid $14.65 per hour, while contracting industries paid $16.92. Conversely, higher paying jobs dominated in the latter 1990s, when industries such as information and professional services, with above-average wages, were expanding. In percentage terms, expanding industries now pay 13% less than contracting ones; in the 1998 to 2000 period, growth industries paid 12% more.” “Nationwide, industries that are gaining jobs relative to industries that are losing jobs pay 21% less annually.1 For the 30 states that have lost jobs since the recession purportedly ended, this is the other shoe dropping—not only have jobs been lost, but in 29 of them the losses have been concentrated in higher paying sectors. And for 19 of the 20 states that have seen some small gain in jobs since the end of the recession, the jobs gained have been disproportionately in lower-paying sectors.” http://www.epinet.org/content.cfm/webfeatures_snapshots_archive_01212004 The drop in participation rate during the recovery is about 0.5% (from 66.4% in January 2002 to 65.9% in February 2004). But where is this number coming from? Clearly this comes from the payroll survey, not the household survey. My gosh! You are all over the place. It is not possible to estimate labor participation rate from the payroll survey. I explained the reason above in detail. They are all from the HH survey! The figures mentioned in the pdf uses Employment-Population Ratio – seasonally adjusted. On this we agree: payroll employment is decreasing as a percent of the population. But your conclusion is that there are millions of huddled, starving masses, while the Heritage Foundation's conclusion -- based upon actual data -- is that there are millions of people working in jobs that the p | ||||||||||
