June 5, 2008

What If ADP’s Employment Report is Right (and Uncle Sam’s Isn’t)?

Specifically, what if it is better at picking up small-business job creation than the government’s Bureau of Labor Statistics (BLS)?

The media isn’t asking this question, even though the two reports have diverged by over 400,000 jobs in the past four months (see latest reports here, here, and here on ADP’s May estimate that 40,000 jobs were added, vs. expectations that it would come in at 30,000 jobs lost — a 70,000 job difference).

So I will.

The differences between business outsourcing behemoth ADP’s National Employment Report and BLS’s Employment Situation Report have been significant since ADP began issuing theirs in roughly April 2006. The report’s preparers, Macroeconomic Advisers, revised the report’s methodology in February 2007.

Since its initial issuance, ADP and BLS have typically differed sharply. It has been easy to chalk this up to the fact that BLS has been at it for decades, while ADP’s effort is new and untested. Perhaps too easy.

ADP’s methodology begins with the payroll data it processes — roughly 1 in 6 employees in the US have their payroll processed by the company. It is also, according to the company, “based on a historical analysis of over 5 years of payroll data and a monthly review of payroll records.”

The government gets to its number of jobs added or lost in any particular month as follows:

Each month State agencies cooperate with BLS, as well as BLS Data Collection Centers, to collect data on employment, hours, and earnings from a sample of about 160,000 businesses and government agencies, which cover approximately 390,000 individual worksites drawn from a sampling frame of over 8 million Unemployment Insurance tax accounts. The active CES sample includes approximately one-third of all nonfarm payroll workers. Sample respondents extract the requested data from their payroll records, which must be maintained for a variety of tax and accounting purposes. Data are collected by telephone, touch-tone self response, computer-assisted interviews, fax technology, voice recognition, and mail. The use of electronic media results in more rapid response times and higher response rates. States also electronically transmit both sample data and geographic estimates to BLS in Washington to speed the estimation and publication processes.

It seems to me that ADP’s methodology has the potential, if done properly, to get closer to being right than the government’s. It should especially be able to pick up changes in small-business employment more quickly.

Here are the differences between the two reports through the first four months of 2008 for private nonfarm payrolls, seasonally adjusted:


ADP shows 117,000 jobs added; BLS shows 312,000 lost. What’s more, the net difference of 429,000 is not far from the sum of job gains by small and medium-sized companies in the service sector, plus the difference between the ADP and BLS in manufacturing jobs lost.

Unfortunately, ADP does not make non-seasonally-adjusted data available, though I have inquired to see if they might consider it. That data, when compared to BLS’s unadjusted numbers, would probably be more valuable. but we’ll work with what’s available.

My theories:

  • ADP is picking up service-sector payroll increases at both new and existing small- and medium-sized service-sector companies more quickly.
  • In manufacturing (a subset of “Goods-producing”), the government is better at quickly picking up job losses, especially big ones, thanks to mass-layoff requirements, than it is typically small individual-employer job gains. ADP’s feedback, based as it is on paychecks actually issued, is more immediate, especially at small companies.

The wild card in all of this is whether ADP has adjusted appropriately for possible differences in the makeup of its client base vs. all employers in the economy when it projects from its in-house payroll data. If they are getting it right, the BLS will have a lot of jobs to add in the coming months — or, if not then, BLS will have to add them in its Comprehensive Annual Revision.

By the way, the BLS’s Comprehensive Annual Revisions in two of the past three years have been huge, so it’s not like anyone, including the recession-obsessed media, can pretend that their problem with picking up new jobs doesn’t exist.

In February 2007′s revision, BLS “found” over 900,000 jobs. The previous year, according to the New York Times, the revision was over 400,000 “found” jobs.

It also happens that a modest increase in total employment so far this year would be consistent with other economic reports we’ve seen, particularly tepid but positive GDP growth, Institute for Supply Management Reports showing overall expansion, decent Orders reports from the Census Bureau, and others.

Regardless of how close the BLS is to ADP’s report tomorrow, “someone” should be digging into this more deeply. Why aren’t they? I’ll suggest that it may partially be because the business press is comfortable in their “of course we’re in a recession” mindset, and the answer to the question I’ve raised here would rock their world.

Cross-posted at NewsBusters.org.



  1. Very interesting! Late last year, Econbrowser’s Jim Hamilton proposed a weighting method between the BLS’ data and ADP’s as a way to get a better “real-time” picture of the employment situation (for which we created this tool), but I don’t think anyone has looked at enough data yet to establish if ADP’s figures are a leading indicator of future BLS job count revisions. If that turns out to be the case, we’re underweighting the jobs counted by ADP in the current edition of the tool (good thing we left the choice of weighting factors up to the user!)

    Comment by Ironman — June 5, 2008 @ 4:49 pm

  2. #1, clever idea, esp the user weighting flexibility.

    Comment by TBlumer — June 5, 2008 @ 5:09 pm

  3. Looking at the ADP numbers, I find it very interesting in the Goods producing sector that the vast majority of jobs are in the small (<50) to medium size (50 to 499) companies with the large (500 and greater) companies only accounting for less than 20% of the work force. The mid size companies employing the largest group of the plurality.

    In the service sector large companies employ less than 16% of the workforce with small companies employing the most.

    These stats certainly do speak to the diversification of the economy away from “Big Business” and ironically show that while big companies have large layoffs such contractions are not as common in smaller businesses who are already very nimble and efficient in the utilization of labor. On the other hand, one could also say just as with the Car Manufacturers, automation or productivity gains outpaceing sales volume rather than an economic downturn is causing unemployment. Big companies starting with a bloated employee roster are intuitively more likely to engage in layoffs as opposed to small companies who are leveraging their employees to the hilt as they increase their sales volume. Large companies sustained the largest losses in the Goods sector while small companies grew (2000 to 2008). The Service sector shows a smaller loss by the big companies however, again the small companies grew.

    Comment by dscott — June 6, 2008 @ 12:23 pm

  4. #3, I didn’t consider the weighting between the categories, but that bolsters the notion that ADP can pick up changes in the hundreds of thousands (or millions?) of smaller companies, esp in services, than BLS can.

    I’d like to see the person from Macroeconomic Advisers get back to me on this, esp given today’s result.

    Comment by TBlumer — June 6, 2008 @ 1:29 pm

  5. ADP’s May estimate that 40,000 jobs were added,

    versus BLS UNadjusted number of 5,000 jobs added for May gives a more reasonable divergance of 35,000 people. Well within a statistical error bar given a Labor Force of approx 146 million people yields a 0.024% variance.

    Comment by dscott — June 6, 2008 @ 2:21 pm

  6. #5, but we don’t know ADP’s unadjusted #s, because they don’t provide them. What they give us is, according to them, seasonally adjusted.

    Comment by TBlumer — June 6, 2008 @ 2:42 pm

  7. [...] BizzyBlog – What If ADP’s Employment Report is Right (and Uncle Sam’s Isn’t)? [...]

    Pingback by Media Mythbusters Blog » Blog Archive » Media Bias Roundup - 06/06/08 — June 6, 2008 @ 4:46 pm

  8. [...] business payroll and other backoffice functions and puts out an employment report of its own. BizzyBlog notes ADP’s numbers have differed from government numbers and wonders if the company’s [...]

    Pingback by More on Today’s Employment Numbers » The American Mind — June 6, 2008 @ 5:13 pm

RSS feed for comments on this post.

Sorry, the comment form is closed at this time.