Over the past 30 years, U.S. real GDP growth in the first quarter has been slower on average than in other quarters. Such an effect is surprising because the U.S. Bureau of Economic Analysis’s statisticians are supposed to remove any regularly occurring intra-year movements in the data through their seasonal adjustment procedures. Indeed, macroeconomists regularly use the adjusted data without paying much attention to the complex methods by which the statisticians remove the seasonal fluctuations. Economic forecasters, in particular, rely upon clean seasonal adjustment procedures to help them obtain informative signals about future business cycle fluctuations and future movements in trend growth. The signals are harder to read when apparent residual seasonal effects — such as the first-quarter effect — infect the data.
In this report, I document the first-quarter effect in the BEA’s estimates for a broad range of components on the product side of the national income and product accounts, provide formal estimates of statistical significance for the first-quarter effect, and show that alternative measures of real GDP based on the work of Aruoba, Diebold, Nalewaik, Schorfheide, and Song (2013) are much less affected by the first- quarter effect than the BEA’s official measure is.
The key findings are:
- From 1985 to 2014, real GDP growth averaged just 1.87 percent (annual rate) in the first quarter, far below the average growth of 3.31 percent in the second quarter, 2.86 percent in the third quarter, and 2.74 percent in the fourth quarter. The differences between first-quarter growth and growth in the remaining quarters are statistically significant. Qualitatively similar results hold when I exclude the last four years from the analysis (2011 to 2014). Notably, the first-quarter effect disappears when I examine the longer sample period from 1959 to 2014, suggesting that the first-quarter effect does not extend as far back 1959.
- A first-quarter effect exists in such components of real GDP as business fixed investment, residential investment, government consumption expenditures, and exports and imports.
- Model-based estimates of real GDP that draw on the BEA’s official measures for real GDP and real gross domestic income (GDI), such as the Federal Reserve Bank of Philadelphia’s GDPplus, display little evidence of a first-quarter effect, suggesting that they provide better readings on first-quarter growth than the BEA’s estimate of real GDP does.
- The absence of a first-quarter effect in the model-based estimates for real GDP carries over to a number of components on the income side of the national accounts, including real gross domestic income itself, suggesting that the first-quarter effect is largely confined to the product side of the U.S. national income and product accounts.
See the April 22, 2015, CNBC report by Steve Liesman in which UBS economist Drew Matus and I participated. Similarly, Goldman Sachs economists Kris Dawsey and Chris Mischaikow expressed concern in a private research note dated April 17, 2015, about potential statistical noise in the BEA’s initial first quarter 2015 real GDP growth release and showed how their proprietary Current Activity Indicator is insulated from first-quarter effects.