Summary
Value-Added Taxes (VATs) raise the most revenue of any tax in OECD countries and some form of VAT has been adopted by almost all countries in the world. While it is commonly assumed that their incidence is borne by consumers, there is limited empirical evidence on this question.
New evidence from Finland analyzes the effect of a large—14 percentage point—VAT decrease in the hairdressing sector in Finland, which was later reversed. The magnitude of the cut, its temporary nature, and the richness of the Finnish administrative dataset allows the authors to track the effect of the VAT cut on prices and firm-level outcomes.
They uncover several new facts about the incidence of VATs. First, prices respond twice as strongly to the VAT increase as they do to the VAT decrease. Second, prices remained higher, even after the VAT cut was repealed, implying that equilibrium prices are history dependent. Third, and relatedly, both markups and profits mirror the response of prices in that they react asymmetrically to the VAT changes. Fourth, the asymmetry depends on firm profit margins: firms with low markups are more likely to pass through tax increases than firms with high markups, while pass-through in the case of tax decreases is homogeneous across firms.
These findings are corroborated through the analysis of a larger sample of all VAT changes that took place between 1996 to 2015 in all member countries of the European Union. Using the impact of these reforms on prices, the authors estimate similar magnitudes of asymmetric pass-through of VAT increases and decreases to prices. Importantly, this asymmetry exists in all the subsamples of commodities, types of VAT changes, countries, and periods considered. This EU-level analysis provides reassurances about external validity issues around the Finnish experiment.
These findings have important policy implications. First, given the magnitude of the asymmetry, policy makers, and in particular, budget scoring organizations, should be careful when using previous pass-through estimates to predict the effect of future VAT changes without accounting for the direction of the change. Second, and relatedly, these findings caution against using VAT cuts during recessions to stimulate demand; such VAT cuts are more likely to stimulate supply rather than demand, since VAT cuts are barely passed through to prices and thus end up benefiting firms. Even more concerning, if VAT cuts are temporary, as was the case for the 2009 UK VAT cut, they could lead to higher equilibrium prices once repealed, further depressing demand.
Explaining the asymmetry is not easy, in part because it is inconsistent with most standard models. The authors suggest that—in certain settings—prices may not be entirely set by market forces alone, and that pricing norms may be strong enough to counteract market forces. Such pricing norms could be driven, for example, by fairness considerations. Indeed, Kahneman et al (1986) show that customers will accept price increases when costs increase and do not feel antagonistic when firms fail to adjust prices downwards when costs decrease. Eyster et al (2020) introduce the insights from Kahneman et al (1986) into a simple monopolistic pricing model, which yields results consistent with several of the empirical findings in this paper.
Main article
Value-Added Taxes (VATs) raise the most revenue of any tax in OECD countries and almost all countries in the world have adopted some form of VAT, with the notable exception of the US. While it is commonly assumed that their incidence is borne by consumers, there is limited empirical evidence on this question. Our research helps uncover an unusual feature of the way their incidence operates: it varies dramatically by whether VATs are increased or decreased. We find that VAT increases tend to be mostly borne by consumers, thus raising consumer prices, while VAT decreases tend to be borne by firms resulting in very little price changes but large profit increases.
Evidence from Finland indicates that prices, and profits, respond more strongly to a VAT increase than a VAT decrease—and that short-term policy changes can have longer-term effects
Our analysis relies on two main approaches. First, we analyze the effect of a large VAT decrease in the hairdressing sector in Finland, which was later reversed. This VAT cut was part of an experimentation program initiated by the Finnish government that was implemented with the explicit goal of assessing the effect of the VAT on economic activity. Because VAT rates are strictly regulated by the European Commission, the choice of sectors on which to experiment was limited to hairdressers and four other small service-sector industries (shoe repair and leather good production, among others). The VAT was cut from 22% to 8% in January 2007, and then reversed in January 2012.
New paper suggests that the prices—and profits—respond more strongly to a VAT increase than a VAT decrease
Importantly, very similar sectors were excluded from this VAT experimentation program. In particular, beauty salons, which in Finland are very distinct from hairdressers, did not qualify for the VAT cut. Taking advantage of this natural variation, we use a standard difference-in-difference framework, by comparing the evolution of hairdressers and beauty salons over time and in response to the VAT changes.
While this change is sector specific and thus potentially lacks some external validity, it offers several other advantages. First, the VAT cut was large, representing a 14-percentage point decrease. Second, the VAT cut was repealed in 2012, thus allowing us to observe a large VAT cut and VAT increase in the same sector and in the same country, ensuring that the asymmetric response of prices to VAT changes is not driven by sector- or country-related differences. Third, the richness of the Finnish administrative dataset allows us to track precisely the effect of the VAT cut on prices and firm-level outcomes.
The Finnish experiment and the underlying Finnish data allow us to uncover several new facts about the incidence of VATs. First, we find that prices respond twice as much to the VAT increase than to the VAT decrease. Second, prices in the treatment group remained higher than in the control group, even after the VAT cut was repealed in 2012. This finding implies that equilibrium prices are history dependent, which is inconsistent with a broad class of standard economic models. Third, and relatedly, we find that both markups and profits mirror the response of prices in that they react asymmetrically to the VAT changes and remain higher in the treatment group even after the VAT is brought back to its original level. Fourth, these asymmetries in mean responses mask important underlying distributional heterogeneity both in prices and in profit margins. The distribution of price changes around the VAT cut is very different from that around the VAT increase. This difference in the two distributions is likely to be due to the fact that firms operating with low profit margins are particularly likely to respond asymmetrically to the VAT changes.
There is also evidence of asymmetric VAT pass-through for the European Union as a whole
Having established these new facts using the Finnish experiment, we turn to a larger sample of VAT reforms which includes all VAT changes that took place between 1996 to 2015 in all member countries of the European Union. These changes virtually span all countries, all commodities and all years included in this period. Using the impact of these reforms on prices, we estimate similar magnitudes of asymmetric pass-through of VAT increases and decreases to prices. Importantly, this asymmetry exists in all the subsamples of commodities, types of VAT changes, countries, and periods we consider.
This analysis allows us to address several limitations of the Finnish experiment and, because of the large number of VAT changes, we can precisely estimate the magnitude of the asymmetry while holding several characteristics of these reforms fixed. For example, the EU-level analysis provides reassurances about external validity issues raised by the fact that the Finnish experiment only applies to a small labor-intensive sector. Furthermore, it allows us to rule out any concerns over the fact that the asymmetry may be specific to the period during which the VAT changes occurred.
Policy Implications
These findings have important policy implications. First, given the magnitude of the asymmetry, policy makers, and in particular, budget scoring organizations, should be careful when using previous pass-through estimates to predict the effect of future VAT changes without accounting for the direction of the change. Indeed, failing to account for the asymmetry would lead to over-estimating the pass-through of a VAT decrease, on average, by a factor of 3.
Second, and relatedly, these findings caution against using VAT cuts during recessions to stimulate demand. This is a policy that was implemented in the UK in 2009, for example. Setting aside the potentially very large fiscal costs of such policies, our findings of asymmetric pass-through suggest that such VAT cuts are more likely to stimulate supply rather than demand, since VAT cuts are barely passed through to prices and thus end up benefiting firms. Moreover, in Benzarti and Carloni (2019), we show that the windfall generated by VAT cuts, at least when considering large VAT cuts in the French restaurant industry, tend to result in higher firm profits rather than higher wages, employment, or investment. Even more concerning, if VAT cuts are temporary, as was the case for the 2009 UK VAT cut, they could lead to higher equilibrium prices once repealed, thus further depressing demand.
Pricing norms may be a key part of the explanation for this asymmetric pass-through
Explaining the asymmetry is not easy, in part because it is inconsistent with most of our standard models. While there are models that can generate some of the findings we uncover, none of which we are aware are able to explain most of the findings. Menu costs coupled with moderate to high inflation can generate some degree of asymmetric pass-through to prices and to profits. Intuitively, firms that face substantial menu costs and high inflation might be reluctant to cut prices when VATs are cut, instead simply holding nominal prices fixed and letting inflation adjust real prices so that, over time, real price adjustments lead to the target pass-through. While such models can generate some degree of asymmetry, they do not match the medium/long term persistence we find for the asymmetry, leading instead to pass-through rates that are symmetric within a few months, depending on the degree of inflation and the size of the VAT change.
New evidence suggests that consumers pay for a VAT increase, but firms benefit when VAT falls
Our best understanding of what could be driving these patterns is that, in certain settings, prices may not be entirely set by market forces alone, and that pricing norms may be strong enough to counteract market forces. Such pricing norms could be driven, for example, by fairness considerations. Indeed, Kahneman et al (1986) show that customers will accept price increases when costs increase but not when demand increases. Conversely, consumers do not feel antagonistic if firms fail to adjust prices downwards when costs decrease. Based on this evidence, Kahneman et al (1986) conclude that “there is a notable asymmetry between the rules of fairness that apply when circumstances increase or decrease the profits of a firm. The rules of fairness evidently permit firms to pass on the entire amount of a cost increase, but […] firms are allowed to retain most of the benefits of a cost reduction.”
More recently, Eyster et al (2020), in an effort to rationalize our finding, show that introducing the insights from Kahneman et al (1986) in a simple monopolistic pricing model yields asymmetric pass-through of taxes. Eyster et al (2020) make two main assumptions. First, customers care about markups: high markups are perceived to be unfair and reduce the utility derived from consuming the good. Second, customers mis-infer markups from prices: they underappreciate the extent to which higher prices reflect higher markups. Firms can educate customers if it is to their advantage—that is, when markups are perceived to be high when in reality, they are low. Eyster et al (2020) show that when costs (or taxes) increase it is more profitable for firms to reveal markups rather than conceal them because perceived markups are likely to increase relative to true markups. When taxes decrease, the opposite holds true: perceived markups are likely to be lower than true markups and firms have no incentive to educate consumers. Eyster et al (2020) show that this asymmetric behavior leads to asymmetric pass-through of taxes. Their model is consistent with several of our empirical findings, including the fact that the asymmetry depends on firm profit margins: firms with low markups are more likely to pass through tax increases than firms with high markups, while pass-through in the case of tax decreases is homogeneous across firms.
This article summarizes ‘What Goes Up May Not Come Down: Asymmetric Incidence of Value-Added Taxes’ by Youssef Benzarti, Dorian Carloni, Jarkko Harju, and Tuomas Kosonen, published in the Journal of Political Economic in December 2020.
Youssef Benzarti is at the University of California, Santa Barbara and NBER. Dorian Carloni works for the Congressional Budget Office. Jarkko Harju is at the VATT Institute for Economic Research in Helsinki, Finland. Tuomas Kosonen is at the Labour Institute for Economic Research in Helsinki, Finland.
