Objects in the economic mirror may not be what they appear

The March release of The Wall Street Journal’s Economic Forecasting Survey showed a sharp downward revision in the forecast for real gross domestic product (GDP) growth in the first quarter (1Q).

In the February release, the average rate of growth predicted for 1Q was 2.0 percent, but in the most recent report, that figure had been revised down to 1.3 percent. The range of economists’ projections for 1Q GDP growth stretched from 3.1 percent down to 0.0 percent in the March survey.

{mosads}My own projection of real GDP growth for the first three months of 2019 was 2.5 percent in my March response to the survey. This was revised down from 3.0 percent in the January survey during which time the average growth rate of respondents was 2.2 percent. The range of projections in January had a span of 3.7-0.5 percent for 1Q.

While I haven’t spoken to any of the other economists on this panel, I suspect our downward revisions were driven by data releases that took place between the January and March surveys. 

Most notably in my mind were:

• an eyebrow-raising retail sales report for December 2018 that showed a 1.6-percent decline in retail sales;
• the advance GDP report for the fourth quarter of 2018 that showed a deceleration of GDP growth to 2.6 percent; and
• the February jobs report that detailed a 20,000 payroll jobs gain that was well below expectations.

Economists tend to be a lugubrious lot, so I must admit that I was not shocked by the results of the March survey, but is the economy really slowing in such a dramatic fashion?

Objects in the mirror may be closer than they appear

Financial markets aside, almost all macroeconomic data are released with a lag and are subject to significant revisions to these initial data releases. These revisions can take place years after the initial estimates of the data are first released.

In order to attempt to divine what the future holds, economists are continually looking backward into this data version of a side-view mirror. What we see however, may not be what it appears.

While convexity is to blame for the distortion of our vehicle’s mirror, complexity is the culprit in the distortion of our initial releases of economic data. Initial estimates contained in the GDP report, labor market report, retail sales and many others are based on partial data often gathered via survey instruments.

Surveys bring with them both sampling and nonsampling errors. Though somewhat complex, they are worth mentioning, particularly given recent events. 

Sampling error is the difference in the results of your survey that arise because of who is included in the survey sample. Inasmuch as the survey sample differs from the entire population you are interested in studying, the survey estimates will not fully reflect the population as a whole. All sample-based surveys, no matter how well-designed, will contain this type of error.

Nonsampling error reflects all other factors that can contribute to the error in the survey estimate. 

From the Census Bureau’s January retail sales release:

“Nonsampling error encompasses all other factors that contribute to the total error of a sample survey estimate. This type of error can occur because of nonresponse, insufficient coverage of the universe of retail businesses, mistakes in the recording and coding of data, and other errors of collection, response, coverage, or processing.”

{mossecondads}While the partial government shutdown will have a very small negative direct impact on GDP growth for 1Q 2019, the fact that data collection agencies are considered non-essential meant that data collection and dissemination by these agencies was significantly impacted, and I believe that the shutdown may very well have increased the nonsampling error of data reports for periods impacted by the shutdown.

It isn’t difficult to imagine that “mistakes in the recording and coding of data, and other errors of collection, response, coverage, or processing” would have been higher as a result of the government shutdown.

That being stated, economists have to forecast with the data they have, not the data they wish they had, and so forecasts for the 1st quarter of the year have been pushed down.

I believe that the data mirror has been fogged over by the effects of the government shutdown and as a result things may be even less like they appear once the final data comes in.

We only have to wait a few years to find out for sure.

Sean Snaith is a professor and the director of the University of Central Florida’s Institute for Economic Forecasting and a nationally recognized economist in the field of business and economic forecasting.