U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
Barbara M. Fraumeni†
Capital stock information is an important component of economic studies examining the relationship between public infrastructure investments and private-sector performance. In order to determine the productivity of public infrastructure, more broadly all forms of public capital, an accurate measure of public capital is needed. A number of studies have examined the productivity of public infrastructure and use capital stock as a primary input, notably those of Aschauer (1989), Munnell (1990), and Nadiri and Mamuneas (1996). There are other studies whose primary focus is the estimation of highway or other types of transportation capital stock, notably those of Faucett and Scheppach (1974), Bell and McGuire (1994, 1997), Dalenberg and Eberts (1994). Unfortunately the capital stock imbedded or constructed in each of these studies and the majority of other studies suffer from at least one of three shortcomings.
A project recently undertaken for the Federal Highway Administration constructs a highway capital stock and describes the underlying concepts and methodologies (Fraumeni, 1999). The three major shortcomings of other studies are as follows:
Very few researchers are aware of the difference between productive and wealth capital stock. Productive capital stock is the stock which has been adjusted for the decline in the potential productive services of an asset as it ages. Wealth capital stock is the capital stock evaluated at its market value. Productive capital stock is clearly the relevant measure for analysis of productivity or the contribution of infrastructure to economic growth. Unfortunately, the majority of previous studies were contaminated by direct use of wealth capital stocks, use of assumptions from wealth capital stock studies, or controlling or benchmarking to wealth capital stock estimates.
Economists favor the light bulb example to explain the difference between the two types of stock. Assume a light bulb is capable of shining for 12 months. At any point in time over that 12 months, until the bulb stops shining, it is 100 % produtive because the intensity of light is constant. However, if one sold the light bulb after 6 months of use, a rational buyer would only be willing to pay approximately half of the original purchase price. In stock measurement, at the 6 month point, the productive capital stock of the light bulb is about double the wealth capital stock.
Until recently, the Bureau of Economic Analysis published estimates of wealth capital stocks (BEA, 1993) which differed from the corresponding productive capital stocks. Wealth capital stocks are needed for the national income and product accounts, which are produced by BEA. Although BEA documentation warns users not to use wealth stocks for productive stocks, this warning apparently was overlooked or unheeded by most researchers.
For components of the stock for which the concepts differ, the Fraumeni productive stock is as much as 1.4 times the corresponding wealth capital stock.1 For most assets, there is no longer a difference between BEA wealth stocks and the corresponding productive stocks (Fraumeni, 1997; Katz and Herman, 1997); however the Fraumeni FHWA study suggests changes in the BEA highway wealth stock methodology that would imply differing estimates for wealth and productive capital stocks.
The distribution of basic components capital outlays for newly constructed or reconstructed roads are clearly different from the distribution of basic components outlays for other types of capital outlays. For example, in the 1997 Cost Allocation Study rural other principal arterials capital outlay for grading as a percentage of outlays for pavement plus grading varied from 6.30 % for resurfacing to 37.6 % for new construction. The Faucett and Scheppach (1974) pavement, grading, and structure split undoubtedly reflects the distribution of basic components outlay for new construction or reconstruction. Approximately one-third of the studies surveyed in Fraumeni (1999) use Faucett and Scheppach (1974) distribution splits. In 1921, new construction and reconstruction was only 8.5% of total capital outlays on locally administered roads, in 1995 the corresponding figure is 10.4% .2 The corresponding figure for state-administered roads, excluding Interstates, is 25.5 % in 1921 and 31.2 % in 1995. Accordingly, new construction and reconstruction splits are inappropriate for most capital outlays.
Some of the components of a highway have a very long service life, for example, commonly grading is assigned an 80-year service life and structures are assigned a 50-year life. Researchers frequently use the perpetual inventory method to calculate capital stocks without an initial year benchmark. The post-World War II stock estimates of only a few studies, notably those of Faucett and Scheppach (1974) and the Bureau of Economic Analysis (1993), are unaffected by this problem as their initial years are as early as the 1870s. Failing to benchmark estimates can significantly downward bias capital stock estimates in studies with an initial capital outlay year in the 20th century. The Fraumeni 1921 benchmarked estimates are 10 times larger than the comparable Bell and McGuire (1994, 1997) and Dalenberg and Eberts (1994) estimates in 1931, their initial year. Although between 1950 and 1960 the difference drops from 0.9 times to 0.08 times larger, the difference persists in the 1970s, 1980s, and 1990s, the Fraumeni estimates varying from 0.08 times to 0.25 times larger than the Bell and McGuire estimates. The danger of underestimation is particularly high for state or local estimates, as the time span for capital outlay series is typically fairly short.
Prior estimates of highway productivity or its contribution to economic growth may be incorrect because of problems in construction of the all-important capital stock input to these studies. The extent of the potential problem does vary by study. The Nadiri and Mamuneas (1996) capital stock estimates are the closest to the Fraumeni estimates of all estimates examined.
†This paper represents views of the author and is not an official position of the Bureau of Economic Analysis or the Department of Commerce. The research described in this paper was conducted while the author was Northeastern University.