Playing with the jobs stats

March 18, 2005

View from Silicon Valley- Playing with the job stats

For the last several months, Silicon Valley has been crowing about its recovery. Several hundred new millionaires were minted via the IPO process. Tax receipts went up (more or less). The stock market (sort of) went up. Housing prices went up (median now up 33%(!) since December, 2004).

The missing piece from Silicon Valley's recovery continues to be net, new jobs. Ignoring the job bubble from the dot-com boom, we are still down -87,000 since 1997 (-99,000 since 1998). Despite all the hoopla and happy noise emanating from our CEO- and political classes, diligent study showed Santa Clara County (SCC) actually lost -5,600 jobs during the twelve months ending December, 2004.

Job news through January was about the same. That is, until the bureaucrats rolled out their new "March, 2004 benchmark."

Long-time readers will recall the Federal Bureau of Labor Statistics (BLS) re-wrote their rules early in 2004. The BLS decided it was reasonable, despite a steadily growing US population, to reduce the total number of households in the US between 2003 and 2004. You don't have to be a statistician to recognize reducing the total population while keeping the same survey or method is like moving in the fences in baseball. Even a flat jobs report will cause a reduction in unemployment, any incremental jobs gains will seem like a homerun. It's like putting the jobs numbers on steroids (sorry, with the grandstanding in Congress this week, I couldn't resist...).

California's Employment Development Department (EDD) recently performed their own maneuver. Suddenly, through the magic of a "benchmark revision," 1.6M lost California jobs disappeared. Ignoring their own statistics for the last four years, the EDD now wants us believe California has been gaining jobs since 2000 without interruption.

Despite these machinations, I was mostly content because Silicon Valley's SCC still puts out their own numbers. With records back to 1990, any changes in trend were easily identifiable and measurable. I suppose this was exactly the problem.

SCC has now changed their methodology. Fortunately, a Google search discovered the original data , through Dec'04, efore it was fully purged from the government's web site.

Old Benchmark:

2004            Labor Force   Employment     Unemployment   Rate
January         875,300       812,900        62,400         7.1%
February        876,700       816,300        60,400        6.9%
March           875,600       815,000        60,600         6.9%
April           867,300       812,500        54,700         6.3%
May             867,000       815,100        51,900        6.0%
June            873,000       818,000        55,000        6.3%
July            879,300       823,100        56,300         6.4%
August          874,900       825,900        49,000         5.6%
September       864,800       816,800        48,000        5.5%
October         868,700       822,400        46,300         5.3%
November        863,500       819,500        44,000         5.1%
December        855,800       816,000        39,800       4.7%

New Benchmark

2005	Labor Force	Employment	Unemployment	Unemp. Rate
January	818,200	768,200	50,000	6.1%
February	827,700	777,400	50,400	6.1%
2004	Labor Force	Employment	Unemployment	Unemp. Rate
January	836,100	771,500	64,600	7.7%
February	835,500	773,800	61,600	7.4%
March	834,600	771,600	63,000	7.6%
April	824,000	768,100	55,900	6.8%
May	821,000	767,600	53,500	6.5%
June	830,200	773,000	57,200	6.9%
July	835,600	778,500	57,200	6.8%
August	832,400	780,800	51,600	6.2%
September	824,800	775,900	49,000	5.9%
October	829,700	780,800	48,900	5.9%
November	827,300	779,000	48,300	5.8%
December	821,400	776,500	44,900	5.5%
Annual Average	829,400	774,800	54,600	6.6%

Under the new regime, there are fewer people and fewer jobs but more unemployed. This is at least gives hope the data may be converging with Silicon Valley's reality.

On the other hand, the varying differences between the two methods gives one pause:

Difference in, as of Jan as of Dec Shift

Labor Force -39,200 -26,400 -32%

Employment -41,400 -39,500 -5%

Unemployment +2,200 +5,100 +132%

A reasonable might expect the relative changes to track one another. If the new benchmark is consistently capturing, or ignoring, a particular category or type of worker, the relative differences between the two surveys should track. However, if SCC is inventing new seasonal adjustments along the way, all bets are off. With "dynamic" seasonal adjustments they can make the data come out pretty much any way they want. (Similar to what seems to already be happening at the BLS and EDD .)

As an aside, the "old" Dec'04 unemployment rate was 4.7%. Under the new benchmark, Dec'04 was 5.5% and Jan'05 was 6.1%. Despite this increase, local politicians were out touting "San Jose tied New York City for the largest year-over-year percentage decline in unemployment in the country," since the new benchmark put Jan'04 employment at 7.7% (a -21% decrease in the unemployment rate). Conveniently forgotten were all the 2004 touts of an unemployment supposedly dropping to 4.7%.

Totally ignored is SCC has 3,200 FEWER jobs while crowing over the Jan'05 unemployment rate decline. If this is "good" news, don't you have to wonder what "bad" news sounds like?

Conclusion:

The last bastion of seemingly-honest numbers appears to now be open to compromise. Take the headlines with a large grain of salt. Better still, visit "SV Stats" at www.viewfromsiliconvalley.com to keep track for yourself.