Johns Hopkins University
The blog on alternative adjustments for seasonal and weather effects in the employment report that I normally maintain will not be posted for January 2019. This is because the weather data have been delayed by the government shutdown. Moreover, the shutdown itself is the biggest complicating factor in understanding tomorrow’s jobs data. The BLS has determined that furloughed federal government employees will be counted as employed in the establishment survey on the grounds that they will eventually be paid. However, they will be counted as unemployed in the household survey.
Even the establishment survey will be impacted by the shutdown. Workers other than federal government employees who did not work in the pay period bracketing the 12th on account of the shutdown will of course not be included in the jobs count. On the other hand, some federal government workers may have secured short-term jobs, and will then be counted in both jobs.
Assuming that there is no further shutdown, comparison of December 2018 and February 2019 employment data will give a better read of the underlying strength of the labor market, and that will be available on March 8.
December jobs numbers
The Bureau of Labor Statistics (BLS) employment report, released on January 4, 2019, shows that 312,000 jobs were gained in December 2018. Each month, I report results from three alternative projections, each of which was calculated using methodology outlined in my past research published in the Brookings Papers on Economic Activity.
You can read more about the methodology below. Here are my alternate projections for December 2018:
- Alternative Seasonal Adjustment (Alt. SA): 309,000 jobs added in December 2018.
- Seasonal and Weather Adjustment (SWA): 326,000 jobs added in December 2018. (Weather effect: -14,000 jobs)
- Alternative Seasonal and Weather Adjustment (Alt. SWA): 333,000 jobs added in December 2018.
Monthly job gains and losses can indicate how the economy is doing once they are corrected to account for the pattern the BLS already expects in a process called seasonal adjustment. The approach for this seasonal adjustment that is presently used by the BLS puts very heavy weight on the current and last two years of data in assessing what are the typical patterns for each month. In my 2013 Brookings Paper “Unseasonal Seasonals?” I argue that a longer window should be used to estimate seasonal effects. I find that using a different seasonal filter, known as the 3×9 filter, produces better results and more accurate forecasts by emphasizing more years of data. The 3×9 filter spreads weight over the most recent six years in estimating seasonal patterns, which makes them more stable over time than the current BLS seasonal adjustment method.
In addition to seasonal effects, abnormal weather can also affect month-to-month fluctuations in job growth. In my 2015 Brookings Paper with Michael Boldin, “Weather-Adjusting Economic Data,” we implement a statistical methodology for adjusting employment data for the effects of deviations in weather from seasonal norms. This is distinct from seasonal adjustment, which only controls for the normal variation in weather across the year. We use several indicators of weather, including temperature and snowfall.
The alternative seasonal and weather adjustment performs the alternative seasonal adjustment and the weather adjustment jointly, adjusting for weather effects while also using a longer data window.
Highlights from past months
Friday, November 2, 2018 — There were two hurricanes in the past two months: Hurricane Florence in September and Hurricane Michael in October. It is important to note that the weather adjustment does not incorporate the effects of hurricanes. Working with national data, Michael Boldin and I found it difficult to estimate hurricane effects with reasonable precision. Thus, we did not include a hurricane factor in our model.
Hurricane Florence lowered employment in September, and as the Carolinas bounced back from this event, it likely boosted the change in employment from September to October. Hurricane Michael lowered the payrolls change from September to October. In terms of relative magnitudes, Hurricane Florence did substantially more damage. But the way that the BLS defines the employment data, a worker is counted as employed for the month if he or she works at any point in the pay period bracketing the 12th. Hurricane Florence made landfall on September 14th, while Hurricane Michael made landfall on October 10th. All else equal, the within-month timing of the two hurricanes makes Hurricane Michael more likely to impact employment in the way that the BLS defines it.
There are a few clues in this morning’s data that suggest the overall direction and magnitude of the hurricane effect in this morning’s data. Absences from work due to weather in the Current Population Survey fell in October to 198,000 from an elevated level in September, though this is still a high number of work absences. Employment in food services and drinking places climbed by 33,000 and is consistent with a sizeable positive hurricane effect. These facts would suggest that bounceback from Hurricane Florence is the dominant effect in this morning’s jobs numbers.
Based on the available information, a reasonable estimate might be that payrolls growth in October was boosted by about 30,000-40,000 from the bounceback from Hurricane Florence net of the negative effects of Hurricane Michael. We will know more when the state and local level data are released on November 17th.
In any case, smoothing through these weather effects, the pace of jobs growth remains very strong. Even with continued recovery in prime-age labor force participation, such a pace of jobs growth will, if continued for much longer, push unemployment down to extremely low levels.
Friday, October 5, 2018 — Hurricane Florence struck the Carolinas in September. It is important to note that the weather adjustment does not incorporate the effects of hurricanes. Working with national data, Michael Boldin and I found it difficult to estimate hurricane effects with reasonable precision. Thus, we did not include a hurricane factor in our model. The fact that employment in food services and drinking places declined by 18,000 is consistent with some hurricane effect. However, the effect of Hurricane Florence on the payrolls total for September is likely to be modest. A worker is counted as employed for the month if he or she works at any point in the pay period bracketing the 12th, and Hurricane Florence did not make landfall until Saturday September 14th. While this could cause some workers to not be employed for the entire previous week, that is likely to be rare. Overall, the jobs report was weaker than the recent trend and that cannot be explained solely or even largely by the hurricane. A reasonable guess based on the available information is that Hurricane Florence lowered jobs growth in September by about 25,000. However, the trend in jobs growth over the last 6 months was not really sustainable. Moreover, month-to-month data are very noisy and so it would be premature to read too much into this morning’s numbers.
Friday, April 6, 2018 — During the month of March, four Nor’easters battered the East Coast in a span of three weeks, two of which (Quinn and Skylar) registered as “notable” or “significant” on the Regional Snowfall Index. This, coupled with colder-than-average temperatures, results in a non-trivial weather effect of –41,000 jobs, resulting in a Seasonal and Weather Adjustment above the BLS Official number. This is the time of year when weather effects in employment data tend to be largest. This estimated weather effect is meaningful, but not especially large. Overall, the March number was notably weaker than the recent trend. Part of this owes to somewhat poor weather, but even after adjusting for this, it represents a slowdown in the pace of employment growth.
Friday March 9, 2018 — According to the Alternative Seasonal Adjustment, the economy added 341,000 jobs in February, even stronger than the BLS Official total of 313,000. The BLS methodology allows a strong month’s data to pull up the estimated seasonal factor substantially. The Alternative Seasonal Adjustment does this to a lesser extent. Hence, the Alternative Seasonal Adjustment shows an even stronger acceleration in seasonally adjusted jobs numbers.
Friday, February 2, 2018 — There was one notable snowstorm in the month of January that could have affected employment—Winter Storm Grayson, which produced blizzard-like conditions in parts of the Northeast, and winter weather advisories as far south as Florida. However, the overall weather effect is small, at +1,000 jobs. Although there was snow in January, this is relatively normal, and temperatures were in line with historical averages. The point of the weather adjustment is to correct for unusual weather. On net, the weather was not at all unusual.
Friday, December 8, 2017 — Today’s estimates for the effects of weather do not take into account hurricanes Harvey and Irma, which clearly had large effects on the previous two jobs reports. The rebuilding effort, coupled with establishments resuming normal operations, likely somewhat boosted jobs growth in November. However, looking at state-level employment numbers for September and October, employment in Texas and Florida more than recouped their September losses in October. Perhaps this is in part because Hurricane Irma ended up tracking farther west than expected. Many establishments may have closed, but not been greatly damaged, and thus resumed normal operations fairly quickly. Therefore, I believe the impact of the hurricanes on the November data was negligible.
Friday, November 3, 2017 — Hurricanes Harvey and Irma substantially impacted the jobs numbers for September. At the time of last month’s jobs report, the BLS reported a loss of 33,000 jobs in September, the first net job loss in more than 5 years. This morning, the BLS has revised September’s number up, from –33,000 to +18,000. Based on state-level jobs numbers that came out on October 20 and other information, including the fact the September jobs numbers were revised upward by BLS, I estimate that these two hurricanes lowered the September employment growth by about 150,000. In past comparable episodes, the jobs numbers have bounced back by about two-thirds in the next month. Hence, I estimate the October employment numbers were inflated by about 100,000 from the bounce-back effect. Adjusting the BLS Official numbers for September and October, the estimated hurricane effect gives an underlying pace of jobs growth of 168,000 for September and 161,000 for October, which is close to the average monthly gains from the previous 12 months. The rebound from the hurricanes and the rebuilding effort should again be a positive factor for jobs growth in next month’s report.
Friday, October 6, 2017 — Hurricane Harvey made landfall in Texas on August 25, and proceeded to flood the city of Houston—America’s 4th largest city—for the next several days. However, because data for the August establishment survey were largely collected before the storm, Harvey had no material effect on August’s employment numbers, but did affect employment for September. Hurricane Irma made landfall in Florida on September 10, which is right around the time of the September establishment survey. Preliminary damage estimates are around $100 billion for Hurricane Harvey and $65 billion for Hurricane Irma, which together puts the damage on par with Hurricane Katrina in 2005. State-level data for initial jobless claims spiked by about 70,000 in Texas and Florida combined. Initial jobless claims have been a reasonable indicator of employment losses for past hurricanes, although it is important to remember that undocumented immigrants can be included in the establishment survey, but cannot claim unemployment insurance. Preliminary estimates by the Federal Reserve Bank of Dallas suggest that Harvey could result in a decline in Texas payrolls of around 40,000 in September, rather an increase of around 30,000 jobs that was otherwise predicted—a 70,000 effect for Texas alone. Employment in Louisiana fell by over 100,000 in the wake of Hurricane Katrina in September 2005, which has resulted in about the same damage as hurricanes Harvey and Irma combined. Taking all this into account, a reasonable estimate of the combined effect of hurricanes Harvey and Irma on September’s employment data is roughly –100,000 jobs, which swamps the typical seasonal and weather effects. Adjusting the BLS Official number with this estimated hurricane effect yields an underlying pace of jobs growth of 67,000. Two final points about the effects of hurricanes on employment numbers must be made. The first is that hurricanes disrupt the physical technology used to collecting the employment data. As a result, future data revisions may be larger than usual. Second is that employment is expected to bounce back as the hurricane cleanup process gets underway, and so we should expect a positive bump for October and November’s jobs data. The bottom line is that September’s is a weak jobs report, but the weakness is exaggerated by the effect of the hurricanes.
Friday, September 1, 2017 — Hurricane Harvey had no discernible effect on the employment data for August. The category 4 hurricane made landfall in Texas on August 26, and proceeded to flood the city of Houston—America’s 4th largest city—for the next several days, causing damage to tens of thousands of homes and forcing many residents to seek shelter. Roads were flooded, and most residents of Houston and the surrounding towns were unable to get to work in the final week of August. However, the establishment survey, which is used to count the number of people employed, was based on data largely collected before the storm. Indeed, the BLS employment release notes that the “establishment survey data collection for this news release was largely completed prior to the storm, and collection rates were within normal ranges nationally and for the affected areas.” Moreover, given the definition of employment in the establishment survey, for the hurricane to affect employment one’s employer would have had to anticipate the hurricane by August 12, which is highly implausible. However, it is likely that the hurricane will affect future jobs numbers, as the economic disruption in the Houston area will be long lasting.
Friday, December 2, 2016 — According to the Alternative Seasonal Adjustment, the economy added 213,000 jobs in November, 35,000 more than the BLS Official total of 178,000. November 2015 was a very strong jobs report and this causes the BLS’s seasonal factor to be moved up. Consequently, this has an echo effect of making the seasonally adjusted numbers weaker in November 2016. My seasonal adjustment averages over more years, and so the echo effect is smaller. This is apparently the reason why my Alternative Seasonal Adjustment gives a somewhat stronger number for November 2016.
Friday, May 6, 2016 — Unseasonable weather can fully explain the weak jobs report for April. The Guardian reported this morning with the headline, “US economy adds just 160,000 jobs in April – further sign of slowdown.” I would argue, however, that the jobs report for April was actually slightly better than expected. Let me explain: April’s jobs gains, released this morning, indicate that the U.S. added only 160,000 new jobs, roughly 40,000 fewer than expected. But this estimate does not account for the effect of unusual weather, which was significant for April. While weather was close to normal in April, it had been milder than normal in each month of the winter. Hence, the effect of weather inflated the level of employment upward through March, and an unwinding of the effect distorted the change in employment from March to April downwards. As a result, there was a sizeable “bounceback” in the data. After adjusting for the effects of unusual weather, employment growth was substantially more robust, at 229,000 jobs. I would argue that all of the slowdown that people are talking about for April is explained by weather. Seasonally adjusted, construction employment was basically flat, a tell-tale sign that weather was a drag on the data.
The author did not receive financial support from any firm or person for this article or from any firm or person with a financial or political interest in this article. He is currently not an officer, director, or board member of any organization with an interest in this article.