US labor market conditions suggest job creation higher than observed in Aug/Sep
Posted on November 4th, 2015
Extracting a common trend
One possibility to summarize labor market conditions is to extract a common trend from several labor market indicators.
Based on previous work from Atlanta Fed (Labor Spider Chart) and from Kansas Fed (Assesing Labor Market Conditions), among others, I’ve compiled a list of labor market indicators: unemployment rate, employment to population ratio, labor force growth, U6 unemployment, short-term unemployment,job losers, unemployment longer than 27 weeks, involuntary part-time, hires, separations, quits, NFIB employment, NFIB plans to hire, Challenger job cuts, jobs plentiful, jobs hard to get, initial claims, Michigan consumer confidence, ISM employment, temp help, average hourly earnings, weekly hours index, weekly payroll index, employment diffusion index, Gallup job creation index.
It is possible to consolidate all the information in the list of labor market indicators in a few orthogonal components using principal component analysis (PCA). The table below shows that the first seven factors explain 92.9% of the data set variance. The first two factors explain 79% of the variance.
The table below illustrates this:
| PC1 | PC2 | PC3 | PC4 | PC5 | PC6 | PC7 | |
|---|---|---|---|---|---|---|---|
| Standard deviation | 4.064 | 2.366 | 1.204 | 0.900 | 0.788 | 0.730 | 0.691 |
| Proportion of Variance | 0.590 | 0.200 | 0.052 | 0.029 | 0.022 | 0.019 | 0.017 |
| Cumulative Proportion | 0.590 | 0.790 | 0.842 | 0.871 | 0.893 | 0.912 | 0.929 |
It is possible, then, to reconstruct the private payroll time series using just the main principal components (PCs). The idea is that the filtered private payroll built this way would embed the common trend among all the labor market variables in the data set, and therefore filter away the idiosyncratic part of the payroll leaving only the true underlying data.
The table below shows the results of a linear regression of private payroll on the first two principal components. One can see that the first two factors are able to explain a high proportion of private payroll variability (R2=0.85).
Adding factors three to seven increases R2 to 0.91. Results can be seen in the table below.
| Private Payroll | ||
| (1) | (2) | |
| Constant | 63.31*** (7.00) | 63.31*** (5.49) |
| PC1 | 50.63*** (1.73) | 50.63*** (1.35) |
| PC2 | 23.60*** (2.96) | 23.60*** (2.32) |
| PC3 | -42.02*** (4.57) | |
| PC4 | -4.16 (6.13) | |
| PC5 | -16.34** (6.99) | |
| PC6 | -20.76*** (7.53) | |
| PC7 | 23.83*** (7.99) | |
| N | 166 | 166 |
| R2 | 0.85 | 0.91 |
| Adjusted R2 | 0.85 | 0.91 |
| Residual Std. Error | 90.16 (df = 163) | 70.70 (df = 158) |
| F Statistic | 461.55*** (df = 2; 163) | 229.75*** (df = 7; 158) |
| Notes: | ***Significant at the 1 percent level. | |
| **Significant at the 5 percent level. | ||
| *Significant at the 10 percent level. | ||
Results for Sep/2015
The chart below plots the actual payroll printed (dots) and the underlying trend using the first two PCs.
The latest private payroll was 118 thousands (Sep/2015), while the underlying trend suggest payroll running at 233 thousands/month.
The underlying trend obtained adding factors three to seven becomes a bit more volatile, but suggest an uderlying trend running at around 194 thousands/month.
Bottom line
- Latest payroll was 118 thousands (Sep/2015).
- The average of the last 3 months was 138 thousands/month.
- The common trend extracted from labor market indicators suggests underlying payroll running at around 194 thousands/month.
Dr. Paulo Gustavo Grahl, CFA (2015-11-04)
