Assessing the importance of direct government expenditures on goods and services [Edits to clear up ambiguity in terminology for readers Salim and Jeff -- MDC 12/18].
One way to approach this issue is to examine the ratio of government spending on goods and services, [otherwise known as government consumption and investment], to GDP in nominal terms. [Hence, this measure excludes transfers]
Figure 1: Ratio of spending on goods and services (all levels), aka government consumption and investment to GDP (blue), and ratio in Ch.09$ (red). NBER defined recession dates shaded gray. Source: BEA, 2013Q3 3rd release, NBER, and author’s calculations.
Notice that in nominal terms, the government consumption and investment spending, i.e., government spending on goods and services, and excluding transfers, ratio is declining. The decline is even more pronounced when expressed using Ch.09$. However, the calculated series is not meaningful as the sum of the components of GDP expressed in real Ch.09$ do not sum to actual GDP, as noted in this post. In this case the approximation to actual using the straight ratio is likely to be poor particularly in the late 1960′s to early 1970′s, as the residual approaches 4% of GDP.
Figure 2: Ratio of residual to GDP, in bn. Ch.09$, SAAR (red). Residual equals actual minus sum of components. Gray bar denotes the reference year. Source: BEA, 2013Q3 3rd release, and author’s calculations.
If the straight ratio of real magnitudes cannot be readily interpreted, given the failure of chain series components to sum up, then what can one do? An alternative is to plot the log ratio. The level of the series has no meaning, but the difference between any two points in series does.
Figure 3: Log ratio of spending on goods and services (all levels) aka government consumption and investment to GDP in Ch.09$ (blue). NBER defined recession dates shaded gray. Source: BEA, 2013Q3 3rd release, NBER, and author’s calculations.
For instance, the difference between the value in 1967Q1 of -1.105 and that in 2013Q3 of -1.695 is 0.59, representing a cumulative 59% change (in log terms). Given the roughly 46 years in the sample, this means that government consumption and investment spending (also known as government spending on goods and services) grew about 1.3% slower than GDP over this period.
Update, 12/18 3:25PM Pacific: Noah Smith speculates that Medicare expenditures are not in the government spending on goods and services series. That guess is correct. See this working paper, page 3. On the other hand, VA hospital expenditures and government health clinics would be included.