Badgering Wisconsin’s Jobs Data

Next month’s election between Wisconsin Republican Governor Scott Walker and Milwaukee Mayor Tom Barrett is arguably the most significant statewide election in 2012.  Governor Walker created a firestorm of controversy when he signed a law limiting the collective bargaining rights of state and local government employees not long after his election in November 2010.  Since then over 900 thousand Wisconsin residents signed a petition for the Governor’s recall.  The economic health of Wisconsin is a key issue in the recall election.  Democrats argue that Republican budget cuts and the corresponding decline in state and local government employment have hurt the state’s economy, while Republicans argue that more fiscal discipline and the promise of lower taxes and regulations will increase Wisconsin’s economic growth rate.

 In the past few days the debate in Wisconsin has shifted to an argument over the reliability of competing jobs data series.  The Democrats point to the widely used establishment payroll survey that estimates monthly employment.  The payroll survey oversamples larger employers but has a difficult time tracking the births of new establishments that are (by definition) not part of the sample frame.

The Republicans, led by Scott Walker, note that the Quarterly Census of Employment and Wages (available since 2001), shows stronger employment growth than the establishment survey.  There is no question that a Census of employers provides more accurate information than a survey.  In fact, the survey data will eventually be revised according to benchmarks provided by the Census.

The problem is that Wisconsin politicians are impatient and the Census data are not available as quickly as the survey data.  The latest official Census figures are only available through September 2011, while the most recent establishment survey is reported through March 2012.  Earlier this week Governor Walker presented preliminary fourth quarter Census data that showed annual employment growth through the end of 2011 while the payroll survey indicated a loss of jobs over the same period.

The following chart, using data from the BLS website, presents comparable non-seasonally adjusted data from the Census and payroll survey.  The employment levels from the Census and establishment survey have a correlation coefficient of .985 over the past decade.  This is not surprising because the Census data are used to benchmark the payroll survey.  Nonetheless, the two series appear to begin to diverge in the second half of 2011.

The next chart illustrates the annual percentage change in the two employment series.  Year over year changes are presented (e.g. the percentage change in employment from January 2001 to January 2002 is reported in January 2012).  The percentage changes in the two series have a correlation coefficient of .99 over the past decade.  The payroll survey indicates that employment growth slowed in the summer of 2011 and turned negative in October 2011.  Employment has declined even further in the first three months of 2012 in the payroll survey.  In contrast, the Census data indicated that employment growth did not decline in the summer of 2011, and according to Governor Walker’s preliminary data, that trend continued throughout the last quarter of 2011.

The two employment series yield conflicting information about employment growth in 2011.  This has generated some political controversy but when the final Census employment data are reported for the last quarter of 2011, they will be a benchmark for revisions of the payroll survey data.  When all revisions have been made the two series are likely to track closely in 2011 as they have throughout the past decade.

It is understandable that political rivals have different views about which policies are more likely to promote growth and prosperity.  It is unusual, however, for candidates to disagree about the reliability of economic data series.  Census data are more reliable than a survey.  The payroll survey provides a valuable, but somewhat noisy, indicator of employment changes that have occurred since the latest Census benchmarks.  Stronger employment growth in the Census data in the last half of 2011 may well be the result of a higher “birth rate” of establishments and start-ups that aren’t included in the payroll survey’s sample.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: