Comments and Discussion

Abstract

Comments and Discussion Janice Eberly and Robert J. Shiller Janice Eberly: Since 1995 the Standard & Poor's 500 index has risen by a factor of three, or about 25 percent per year. Impressive as this performance is, it is modest in comparison with the sevenfold, or 45 percent per year, increase logged over the same period by the S&P High Technology Index as of the date of this conference. This astounding performance suggests to many that an asset bubble has developed in equity markets. Alternatively, the value of firms' intangible capital may have risen substantially, accounting for at least part of the spectacular increase in their market values, particularly in "new economy" industries. Stephen Bond and Jason Cummins attempt to sort out these two explanations empirically. This is an admirable and difficult task, since neither the market values of intangible assets nor asset bubbles are directly observable. Lacking direct measures, the authors construct a model and use proxies for the desired data and argue that intangible and bubble effects can be inferred from the results. The starting point of Bond and Cummins' argument is a measure of firms' value based on analysts' earnings forecasts. This is used to construct a measure of average q based on expected earnings rather than markets' valuation of those earnings. To tackle the problem of unobserved intangible capital, the authors then set up a model that allows them to use only the flow of investment in, rather than the stock of, intangibles. They estimate this model as a linear regression of physical capital investment on their constructed measure of fundamentals and intangible investment. An asset-based measure of fundamentals, Tobin's q, is added to this regression as a diagnostic: if the "fundamental measure" is correctly capturing investment incentives, then any additional value in Tobin's q is simply noise, or a [End Page 109] bubble.1 Such a bubble, they argue, should not affect investment and therefore should not be significant in the regression. The final test is to analyze new economy firms separately. Even if the measure of intangibles is not perfect or does not apply to all firms, then by dividing the sample, it should be possible to ascertain whether intangibles are relatively more important in the new economy. Based on this empirical strategy, the paper finds that the asset-based measure of Tobin's q does not provide incremental information in the investment equation, and that the effect of intangibles in these equations is modest—in both old and new economy firms. The paper concludes that the explanation for equity prices based on intangibles is a "fiction" and that equity prices instead contain a large bubble component—"noisy share prices"—that is pervasive across firms and industries and over time. The approach the authors use to reach these conclusions has four necessary ingredients: the measure of fundamentals, the use of the investment equation to identify the fundamental part of the firm's value, the measure and specification of the effects of intangibles, and the identification of the new economy. Each of these links deserves careful attention. To construct their measure of fundamentals, Bond and Cummins use data from equity analysts who forecast earnings at the firm level. The authors gather two years of earnings forecasts and project a "long run" growth rate that goes out to year five. To calculate a firm's value based on these three pieces of data, the authors have to make a number of assumptions. For example, they assume that earnings after year five will grow at a fixed, economy-wide rate of 6 percent, and they discount the future earnings of all firms at the same rate. This common discount rate is the current-year long-term Treasury bond rate plus a constant risk adjustment of 8 percent. Thus, all firms implicitly have a beta of one and assume a fixed interest rate over a five-year horizon. In addition, by discounting forecasted earnings rather than cash flows, the valuation incorrectly allows for depreciation but not investment expenditure. Although all of these measurement problems would be likely culprits if the resulting valuation were not empirically relevant, the results suggest instead that it is strongly...