_________________________________________________________________________________________________________
Di g i t a l Capi t a l a n d S u p er st ar F i r m s 4
HUT C HI N S CE N T ER ON FI S C A L & M ONE T AR Y P O L I CY A T B R O OK I N GS
Prior work has reported correlations between IT investments, firms’ market value, and the risk-reward
profile of firms’ traded assets. This tells us something important about the value and pricing of these
assets (Bharadwaj et al., 1999; Brynjolfsson et al., 2002; Dewan et al., 2007). The value the market
assigns to these assets, however, cannot tell us about their productivity or contribution to economic
growth. By using Hall’s approach with firm-level data, this paper is able to directly tie the flow of services
that firms derive from their digital intangible capital to economic outcomes such as productivity. The
identification of these flows and their economic properties can then be used to explain productive
differences across firms. The ability to measure digital capital also creates further possibilities to measure
the implications of digital capital accumulation for important macroeconomic trends.
This work builds on prior research connecting IT investments to differences in the productivity of US
firms (Dewan and Min, 1997; Stiroh, 2002; Brynjolfsson and Hitt, 2003), and the important role in this
relationship played by complementary organizational investments (Bresnahan et al., 2002). In contrast to
these latter studies, we measure the quantity of intangible capital directly rather than making inferences
about its presence through observing correlations between business practices and IT investment,
productivity, or market value. While this approach cannot identify the specific business practices that
make up digital capital, it has the advantage of creating an aggregate measure of the overall stock of
digital capital without having to identify and measure all of its constituent investments. Once this stock is
identified, it becomes possible to study the properties of digital capital and the statistical distribution of
these digital capital assets across firms and over time. As noted earlier, the separation of price and
quantity of digital capital also enables a distinction between digital capital-related rents and increases in
productive capacity.
Our approach reveals four important facts about the role of digital capital in explaining features of the
modern economy. First, the market value of digital capital rose sharply during the late 1990’s but then fell
in the early 2000’s, reflecting changes in the price of digital capital during the dot-com boom and
subsequent bust.
Second, the value of digital capital began to rise again from 2010 onward, with the timing coinciding
with a wave of innovations based on mobile technologies, cloud computing, big data, data science, and
most recently, artificial intelligence (AI). With the important exception of the period corresponding with
the dot-com boom and subsequent bust, the long run increases in value can largely be attributed to
changes in digital capital quantities, rather than prices. Quantities of digital capital rise fairly steadily and
substantially over the course of our nearly thirty-year panel. By the end of the panel, digital capital
quantities account for 20-25% of the levels of physical capital for firms in our sample.
Third, there is substantial heterogeneity among firms in terms of quantities of digital capital they
own. The majority of the increase in quantities concentrated in a subset of superstar firms, which we
define as those within the top decile of our sample in terms of market value. This concentration can be
observed not only in terms of aggregate digital capital accumulation, but also for the development of
digital capital on a per-employee basis, and it contrasts with patterns of accumulation of other assets such
as those classified as property, plant, and equipment. Furthermore, heterogeneity has been increasing
during the course of our panel, as the top firms have pulled increasingly further away from the rest.
Fourth, by creating firm-level measures of digital capital quantities, we can estimate how the
accumulation of this form of capital contributes to productivity and growth. In productivity regressions,
we estimate the contributions of our digital capital quantity measures alongside IT capital stock measures.
We find that the contribution of digital capital to growth during this period was approximately double that
of IT capital stock. Moreover, current year levels of digital capital are most predictive of productivity levels