Does digitalisation make the world more unequal? Economic research so far seems to suggest yes. There is substantial evidence that digitalisation leads to an increase in income inequality throughout the Western world (Akerman et al. 2015, Gaggl and Wright 2017, Burstein et al. 2019).1 As the literature shows, information and communication technology (ICT) tends to make high-skilled workers more productive, whereas low-skilled workers are often made redundant. As a result, the skill premium rises.
We thus know a lot about digitalisation and incomes by now. From a welfare perspective, however, we care more about consumption. Digitalisation affects consumption inequality not only through income changes, but also through changes in consumer prices. A priori, it is unclear whether this price effect works in the same or opposite direction from the income effect. If the increased use of digital technology makes some consumption goods cheaper than others, it will benefit the income groups that consume relatively more of these goods. Depending on what those goods are, either rich or poor households could be the beneficiaries.
In a new paper (Arvai and Mann 2021), we study the effect of digitalisation on consumption inequality, quantifying both the income and price effect. We use US household data to document that the price effect works in the same direction as the income effect, favouring high-income households. In a two-sector growth model, we show that the price channel accounts for 22.5% of the increase in consumption inequality between 1960 and 2017. This implies that digitalisation has more uneven effects than the increase in income inequality suggests.
Digitalisation and consumption inequality in the data
To study how households differ in their consumption of digital products, we develop a new measure of the ICT share of consumption goods. Our approach focuses on the capital stock used in production, and we study to what extent an industry uses digital assets, like computer software and hardware or intellectual property products, rather than non-digital equipment and structures like production plants.
Assembling data across 61 industries between 1960 and 2017, we find that the average share of ICT capital in the overall capital stock has increased from virtually 0% in 1960 to 16% in 2017, with large variation across industries. We account for the digitalisation content of intermediate products by relying on the input-output structure of the production network and derive a digitalisation measure of final commodities.
We link this measure to the US Consumer Expenditure Survey, an annual survey on the spending patterns of American households, and in this way compute the overall ICT share of consumption of individual households. Figure 1 shows which categories matter for high- and low-income households and how they differ in their ICT intensity. Low-income households disproportionately consume food and textiles, which are produced using a low ICT share. In contrast, categories that tend to be more important for rich households, such as finance and insurance or education, have higher ICT shares. We also document that consumer price inflation has been weaker for ICT-intensive commodities, which means that digitalisation benefits consumers of ICT-intensive goods.
Figure 1 ICT intensity and relative expenditure shares by commodity in 2017
Notes: Each circle corresponds to a commodity. The size of the circle reflects the share of expenditures for that category of the median household. The x-axis shows the log ratio of the expenditure share of the top 10% relative to the bottom 10%. The more to the right, the more important a category is for the rich. Values around zero indicate that a category is equally important to rich and poor. The y-axis shows the ICT intensity of the commodity, the horizontal green line indicates average ICT intensity. Data are for 2017. \
Source: BEA, CEX, and own calculations.
Figure 2 shows our comprehensive measure of ICT intensity of consumption for households at different percentiles of the income distribution. Households in the tenth percentile have a 13% higher digital share in their consumption basket than households in the first percentile, and the difference has become more pronounced over time.
Figure 2 ICT intensity along the income distribution
Notes: The graph shows the ICT share of the consumption basket by percentile for different sub-periods.
Source: BEA, CEX, and own calculations.
A simple model to quantify the income and price channel
To find out how important changes in prices and changes in incomes are for consumption inequality, we construct a two-sector growth model. Output of both sectors is produced using two types of capital, ICT capital and non-ICT capital. Sector 2 uses ICT capital more intensively than sector 1. The economy is populated by two types of agents that differ by skill endowment. High-skill labour is complementary to ICT capital, whereas a composite good constructed from these two inputs is substitutable to low-skill labour. Digitalisation is modelled as an increase in the rate of transformation of output into ICT capital (Karabarbounis and Neiman 2014). The ICT-intensive sector 2 benefits relatively more from this technology trend, which means that the relative price of sector 1 increases. At the same time, the skill premium increases. In a setting with non-homothetic preferences, the effect of changing relative prices depends on the agent's position in the income distribution. In line with our empirical findings, we assume that the ICT-intensive good is the luxury good, which is consumed more intensely by the high-skill, high-earner households.
We calibrate the model to the US economy between 1960 and 2017 such that we match both the increase in the skill premium and the increase in the relative price of non-ICT goods. Consumption inequality increases by 18.1%. To express this number in monetary units, we do a compensatory variation, asking how much additional income we need to give to households in 2017 in order to compensate them for the change in the relative price. Then we compare this number to the actual income increase. The left panel of Figure 3 shows that high-income households benefit both by requiring a lower level of compensatory income and by receiving a larger increase in actual income. They experience an increase in welfare that is equivalent to 22.3% of their initial income, while low-skill households only experience an increase of 5.3%.
To assess the relevance of the income and price channel, we carry out a counterfactual exercise. We quantify how much the price channel contributes to consumption inequality by comparing consumption in the baseline model to a version with fixed relative prices. The right panel of Figure 3 shows that without the relative price change, consumption inequality would have increased by 22.5% less. So, it is important to take the price channel into account. In providing a first estimate of the size of this channel, our paper offers a starting point for more follow-up research on the effect of digitalisation on welfare and inequality.
Figure 3 Welfare decompostion
Notes: The left panel shows how much income households get with the compensatory variation (blue) and how much income households actually get in 2017 (red) relative to the initial 1960 income. The right panel shows relative consumption in the baseline model (red solid line) and in a counterfactual exercise where prices remain constant (dashed blue line).
Acemoglu, D and D Autor (2011), “Skills, tasks and technologies: Implications for employment and earnings”, in Handbook of Labor Economics 4: 1043–1171. Elsevier.
Akerman, A, I Gaarder, and M Mogstad (2015), “The skill complementarity of broadband internet”, The Quarterly Journal of Economics 130(4): 1781-1824.
Arvai, K and K Mann (2021), “Consumption inequality in the digital age”, SSRN Working Paper.
Autor, D H, L F Katz, and M S Kearney (2008), “Trends in US wage inequality: Revising the revisionists”, Review of Economics and Statistics 90(2): 300–323.
Burstein, A, E Morales, and J Vogel (2019), “Changes in between-group inequality: Computers, occupations, and international trade”, American Economic Journal: Macroeconomics 11(2): 348-400.
Gaggl, P and G C Wright (2015), “A short-run view of what computers do: Evidence from a UK tax incentive”, VoxEU.org, 20 Aug.
Hémous, D and M Olsen. (2020), “The rise of the machines: Automation, horizontal innovation and income inequality”, American Economic Journal: Macroeconomics (forthcoming).
Karabarbounis, L and B Neiman (2014), “Labour shares, inequality, and the relative price of capital”, VoxEU.org, 15 Nov.
1 When considering automation more broadly, evidence is even more ample (e.g. Autor et al. 2008, Acemoglu and Autor 2011, Hemous and Olsen 2020).