Swing Trading Strategies & Stock Picks Since 2003

I thought I'd add a few observations on the latest employment report (other reports here: [NYT], [WSJ RTE/Izzo] [CR], [Economist's View]). First, by an alternate measure, employment is improving more rapidly than the standard nonfarm employment (NFP) measure. Second, the alternate measure increased faster than nonfarm payroll employment over the period of temporary Census hiring. Third, aggregate hours worked in the private sector continues to rise faster than private sector employment. Fourth, the NFP growth consistent with zero GDP growth is lower in the last decade, versus previous decades, even while the elasticity of NFP growth with respect to GDP growth has risen.

The Standard and Nonstandard Time Series

Figure 1 illustrates nonfarm payroll employment, and the same series excluding temporary Census workers.

Figure 1: Nonfarm payroll employment (blue line), NFP excluding temporary Census workers (red line), and civilian employment series adjusted to conform to NFP concept (green line), all seasonally adjusted. Gray shaded area indicates recession, assuming trough at 2009M06. Source: BLS via FRED, NBER and authors calculations.

As described in earlier posts [1] [2], the adjusted civilian employment series is constructed to conform to the nonfarm payroll series concept, but based on data from the household survey. One characteristic of this research series is that it has higher variability, measured as the standard deviation of the log first difference, than the nonfarm payroll series, so one might want to downweight its importance relative to the NFP measure.

Interestingly, the adjusted civilian employment series shows a bigger jump than the NFP series. The elasticity of adjusted civilian growth with respect to nonfarm payroll employment growth over the 2009M12-10M08 period is 1.3. Over the entire 2005M01-10M08 period, the elasticity is 0.88. It could be that the extra spending associated with the extra Census workers induced greater employment in firms not well covered by the establishment survey; or it could just be an artifact of measurement error.

Employment versus Hours Worked

Figure 2 illustrates the nonfarm payroll series, the (log) private payroll series, and aggregate hours index for the private sector, all normalized to zero at 2007M12.

Figure 2: Log nonfarm payroll employment, seasonally adjusted (blue line), private sector employment (red line), private sector aggregate weekly hours index (green line), all seasonally adjusted, normalized to 0 at 2007M12. Gray shaded area indicates recession, assuming trough at 2009M06. Source: BLS via FRED, NBER and authors calculations.

Private aggregate hours are rising more rapidly than (private sector) employment (this characterization remains intact from last month's post on the employment situation.

Pedagogical Moment / Employment-Growth Stylized Facts

In comments to a recent post, Econbrowser reader CoRev wonders if an increase in GDP necessarily implies an increase in employment (this is in the context of whether an increase in GDP due to ARRA necessarily induces an increase in employment). Most theories I know suggest this is true (although, I suppose that if the elasticity of substitution between labor and capital were sufficiently high, one could get inverse movements between the two series). But this is an empirical question in the end. So here is a slight digression on the subject of how linked are GDP and nonfarm payroll employment. Figure 3 below shows a scatterplot of the first log difference of NFP against first log difference of real GDP. The blue dots pertain to 1967Q1-85Q4; red dots to 1986Q1-2007Q3; and green dots to 2007Q4 to 2010Q2. Interestingly the (short run) elasticity of employment growth with respect to GDP growth is slightly higher in the most recent period compared to the 1967Q1-85Q4 period.

Figure 3: Log first difference of nonfarm payroll employment against log first difference of real GDP (both seasonally adjusted); blue circles, 1967Q1-85Q4; red circles, 1986q1-07Q3; green squares, 2007Q4-10Q2. All monthly NFP figures converted to monthly by arithmetic averaging. Source: BLS via FRED II, BEA 2010Q2 2nd release, and author's calculations.

There is a clear positive correlation, with a slope coefficient of 0.4 for the 1967-85 period, to about 0.5 for the 1986-2010 period (statistically significant at the conventional levels).

What changes is the intercept in the regression of first differences on first differences; it's lower in more recent periods. A similar pattern is observed with respect to year-on-year changes. In other words, at zero GDP growth, employment growth is now lower (essentially 0%, post 1985, -0.2% post 1999) than it was in the 1967Q1-85Q4 period (0.23% per quarter, all figures not annualized).

A more formal (statistically speaking) treatment of the link between employment and GDP is in this post. Hence, for reader CoRev's benefit: (1) there is a short run correlation between GDP growth and employment growth; there is a long run relationship (in a statistical sense, called cointegration) between log GDP and log nonfarm payroll employment. If GDP is increased due to say an increase in government spending on goods and services, then in general nonfarm payroll employment will also increase.