Friday, June 20, 2014

Tradeoffs Between Timeliness and Accuracy

Carrie Mayne, Chief Economist

Adding up the 1.3 million or so jobs in the Utah economy may seem simple but in reality takes time.  To compile this information, analysts follow the same method used by every state in the country, which is defined by the U.S. Bureau of Labor Statistics (BLS). This method involves using quarterly reports submitted to the state’s unemployment insurance system, which accounts for the vast majority of jobs counts in the state.  This data collection process takes about four months after the end of that given quarter to complete, giving us a very accurate picture of the job market.  For example, we won’t know until November exactly how many jobs there are in our state today (June).

While a complete and accurate count is ideal, the cost of waiting several months to get the information  is not. BLS understands this and therefore conducts a more timely employment survey every month.  The survey includes a representative sample of employers across the state and various industries who are asked report their monthly employment.  Estimates of total employment are then calculated based on a model defined by BLS.  How well the estimates will reflect the “true” employment levels depends on:  1. How well the job creation of the sampled employers reflects the employers they are meant to represent, and 2. The accuracy of the model intended to explain the broader statewide jobs picture.

Each month when we report the job growth rate estimated from the monthly survey, we compare the estimate to all the other information we are studying to understand the current state of the economy and the trends that are driving economic activity.  Sometimes the survey lines up with our expectations. Other times we speculate that the monthly numbers either over- or under-estimate current activity.

Click to enlarge
So, what do we think about the latest numbers?  The accompanying graph shows the last twelve monthly survey job estimates along with the most recent six months of complete count data (the data that takes longer to produce).  Over this period, it seems the survey has been underestimating the job count by an average of about 2,600.   Does that mean the current estimates are also underestimating?  The unemployment rate, which has trended downward for quite some time, may indicate that job growth is actually stronger than what was estimated from the survey.

Only time will tell if our conjectures ring true.

No comments:

Post a Comment

We welcome your comments! However, we do not post comments with "links" or those comments directing to your website. These will not be published. Thank you