Productivity Puzzles - Intangibles, Investment, Labor Quality
Monday, Jan. 4, 2021 3:45 PM - 5:45 PM (EST)
- Chair: Louise Sheiner, Brookings Institution
Mortgage Borrowing and the Boom-Bust Cycle in Consumption and Residential Investment
AbstractThis paper studies the transmission of the major shocks in the U.S. housing market in the 2000s to consumption and residential investment. Using geographically disaggregated data, I show that residential investment is more responsive to these shocks than consumption, as measured by elasticities and the implied contributions to GDP growth. I develop a structural life-cycle model featuring multiple types of housing investment to understand the large responses of residential investment. Consistent with the microdata, the model generates lumpy debt accumulation, lumpy housing investment and a strong correlation between mortgage borrowing and housing investment at the early stage of the life cycle. In the model, households move up the property ladder by increasing their mortgage debt after they have accumulated enough home equity. Since liquidity constraints and fixed costs prevent especially young homeowners from acquiring their desired home, shocks to their borrowing capacity have a large impact on residential investment.
Reliability Statistics for Quarterly Labor Productivity Estimates
AbstractThe U.S. Bureau of Labor Statistics (BLS) produces estimates of quarterly labor productivity growth for the nonfarm business sector. BLS produces a preliminary estimate and two revised estimates for each quarter. Productivity estimates continue to be updated as underlying data on both output and hours worked are revised further after that time. This paper examines revisions to estimates of productivity growth and the underlying data series used to construct it, and constructs intervals for the later estimates.
The BLS’s Labor Productivity and Costs (LPC) program produces quarterly estimates of labor productivity growth. For these estimates, BLS combines output data from the Bureau of Economic Analysis (BEA) with employment and hours data compiled from three BLS surveys: the Current Employment Statistics (CES) survey, the Current Population Survey (CPS), and the National Compensation Survey (NCS). Of these data sources, two—the BEA output data and the CES employment and hours data—are revised multiple times after they are first released. For each reference quarter, BLS releases three regularly scheduled estimates of labor productivity growth. The first estimate (prelim) is issued within 40 days of the end of the reference quarter. This initial estimate is revised as new data become available. The first revised estimate (R1) is released 30 days after prelim, and the second revised estimate (R2) comes out 60 days after that. R2 is the last regularly scheduled release covering the reference period.
For this study, we develop intervals based on revisions under the maintained assumption that each revision moves the estimate closer to the true value. Our focus is on prelim-to-R2 and R1-to-R2 revisions. We begin by summarizing the revisions and examining factors that might affect the size and direction of the revisions. We then consider alternative ways of generating reliability estimates for the quarterly estimates of aggregate U.S. nonfarm labor productivity growth. Our main focus is on estimates of growth from the previous quarter, since they receive the most attention. We report results for three alternative methods for constructing intervals. The first is the modified confidence interval methodology discussed in Fixler, et al, (2014) and Fixler, et al (2018). The second is model-based, which allows us to control for differences across quarters. And the third method constructs intervals based on percentiles of historical revisions, similar to the reliability estimates for the Fed’s IPI. For each method, we generate 70-, 80- and 90-percent intervals.
Labor Composition and Quality Using Augmented CPS Data On Industry and Occupation
AbstractThe U.S. Current Population Survey (CPS) classifies the jobs of respondents into hundreds of detailed industry and occupation categories. The classification systems change periodically, creating breaks in time series. Standard concordances bridge the periods, but often leave empty cells or artificially sharp changes in time series. For estimates about the composition of the workforce by industry, researchers want smoother time series for industry and occupation.
We used training data sets in which industry and occupation have been classified by specialists into two industry and occupation category systems – that is, they are dual-coded.
For each employed CPS respondent in recent decades we apply prediction methodologies, principally random forests, to impute standardized industry, occupation, and related variables. The imputations use micro data about each individual and large training data sets about the population. Augmented data sets of this kind can serve research on many topics. We test the industries and occupations in the resulting augmented data sets for smooth population proportions and wage levels and for how well they match known trends, benchmarks, and alternative data sources. We have experimentally applied the same techniques to American Community Survey data to further add to the CPS with estimates from this larger but less frequent data source.
We apply the resulting “augmented CPS” data to create labor composition indexes that track the overall education and experience of the workforce in each industry, using the standardized imputed industry. The resulting indexes are more stable than those from the original CPS, which makes it more feasible to cover smaller industries.
Case Western Reserve University
Federal Reserve Board
Central University of Finance and Economics
U.S. Bureau of Economic Analysis
- O0 - General
- E0 - General