The changing composition of output and the great moderation

Abstract

An additional explanation is provided for the decline in output variability that began in the mid-1980s. Using state, regional and aggregate data for the US, we examine the shifting influence from manufacturing to services on this variability. At all levels, we find support for this output composition change contributing to the reduced variability of output growth.

Notes

¹ Summers (2005) argues that because the timing of the decline in output growth variability was not synchronized across countries this weakens the good luck explanation, while strengthens the other two.

² The BEA has similar data for the 1997 to 2005 period. However, that data was constructed differently from the 1977 to 1997 BEA data. Rather than complicating matters unnecessarily (by attempting to alter one series to conform with the other), we simply used the 1977 to 1997 series because it included the relevant period – the mid-1980s (Figs. 1 and 2).

Fig. 1 Time-series values of manufacturing share of output and service share of output for the US economy over the 1977 to 1997 period

Fig. 2 Time series values of manufacturing share of output and service share of output for the eight Bureau of Economic Analysis regions over the 1977 to 1997 period

³ The timing of the switch of service's share of output dominating manufacturing was not synchronized across regions. This is consistent with Summers (2005) observations of industrialized countries.

Table 1. State-level OLS estimates from Equation 1 of RATIO 's coefficient (i.e. α₁). In this case, i in Equation 1 represents the 50 US states plus the District of Columbia. States are listed by their BEA defined region (via standard state abbreviations)

⁴ Employing a similar fixed-effect procedure for a regression pooling the 51 states yielded a coefficient for RATIO of 0.06 for the aggregate US economy.