Conference Agenda

Capital incentive policies are widely used to stimulate investment and economic growth by lowering the cost of capital. These policies often implemented through tax allowances, subsidies, or accelerated depreciation have historically been designed to encourage investments in tangible assets such as machinery, equipment, and information technology (IT) hardware. However, the emergence of cloud computing has fundamentally changed how firms acquire digital technologies. Rather than purchasing servers and software as capital investments, firms can now access computing power, storage, and software through cloud providers as digital services. This shift raises an important policy question: do traditional capital incentive policies distort firms’ adoption of modern digital technologies that are accessed as services rather than capital goods?

This paper examines whether capital incentive policies affect the diffusion of cloud computing and related data technologies such as big data analytics and artificial intelligence (AI). We focus on the United Kingdom’s Annual Investment Allowance (AIA), a tax policy that allows firms to deduct the cost of capital investment from taxable profits up to a specified threshold. Because cloud expenditures can already be fully expensed in the year they occur, while capital investments typically cannot, the introduction of the AIA effectively reduced the tax advantage of cloud services by providing similar tax treatment to investments in physical IT capital. This creates a setting in which tax policy may influence firms’ choice between investing in on-premise IT infrastructure and adopting cloud-based technologies.

To examine this question, we exploit changes in the AIA investment thresholds as a quasi-natural experiment. These threshold changes altered the marginal incentive to invest in capital for some firms but not others. Using this variation, we implement a difference-in-differences empirical strategy comparing firms whose historical investment levels placed them below newly introduced thresholds (treated firms) with those whose investment levels remained above them (control firms). Our analysis combines several novel datasets from the UK Office for National Statistics, including firm-level panel data on cloud, big data, and AI adoption, detailed firm investment data, and matched employer-employee records that allow us to examine changes in demand for data-analytics workers.

Consistent with the policy’s objective, we find that the AIA significantly increased firms’ investment in physical capital. Firms eligible for the allowance increased total capital investment by approximately 62 percent relative to comparable firms that were not affected by the policy. However, this increase in capital investment came at the expense of adopting newer digital technologies. Firms exposed to the AIA were 17 percentage points less likely to adopt cloud computing, a substantial effect relative to the average cloud adoption rate of roughly 28 percent in our sample.

The negative effects are particularly pronounced for cloud services related to data storage and processing—functions that most closely substitute for in-house IT infrastructure. Moreover, the policy also slowed the adoption of complementary digital technologies. We estimate that treated firms were 18 percentage points less likely to adopt big data analytics and 3 percentage points less likely to adopt AI. Aggregate calculations suggest that the policy reduced overall cloud adoption in the UK by roughly 7–9 percentage points, delaying cloud diffusion by more than one year. Similarly, the diffusion of big data and AI appears to have been delayed by approximately one to two years relative to a counterfactual without the policy.

To better understand the organizational implications of these technology choices, we examine worker-level outcomes using matched employer-employee data. If the adoption of data technologies increases demand for data-analytics skills, a slowdown in adoption should be reflected in labor market outcomes. Consistent with this mechanism, we find that firms affected by the AIA reduced demand for workers performing data-analytics tasks, with wages for these workers falling by about 1 percent relative to comparable workers in untreated firms. Importantly, we find no comparable effects on wages for other types of workers, suggesting that the policy specifically reduced demand for skills associated with data-driven technologies rather than labor demand more broadly.

Our findings contribute to several literatures. First, we provide new evidence on how policy can shape the direction of technological change, highlighting that policies designed for earlier technological paradigms may inadvertently slow the diffusion of newer technologies. Second, we add to the growing literature on the diffusion of data technologies, showing that cloud infrastructure plays a critical role in enabling the adoption of big data and AI. Finally, we extend the literature on tax incentives and firm investment by demonstrating that while capital incentives successfully stimulate investment, they may generate unintended consequences for the adoption of emerging digital technologies.