This paper is the first to analyze welfare impacts of data center support in a general equilibrium setting. We examine the benefits and costs of an electricity tax exemption for data centers in Sweden, effective 2017-2023. The Government motivated the exemption by competitiveness concerns and the high electricity use in such centers.

The study show that the net benefits of this policy are closely related to the value of a tax elasticity; it is a sufficient statistic for evaluating the policy’s efficacy in general equilibrium. This parameter summarizes a range of economic forces such as: crowding out of existing firms, increasing profits in the electricity sector, effects on electricity tax revenue, the inability to tax profits of data centers owned by large multinational and so on and so forth.

We take our model to the data by combing panel data from the Tax Authority and several other sources on individual establishments (N=135, 2008-2020). We also use a time series on sectoral data. The time series analysis suggests a low tax elasticity, significantly lower than the threshold value of 1 in absolute value. The casual effects analysis using the panel data shows significant heterogeneity regarding the effect of the reform on electricity use. It invariably suggests that the treatment effect is positive; reducing the tax increased electricity use among those granted.

Lessons learned include the oft-repeated plea for evaluation planning: before a policy is put in place, a plan for evaluating the reform should be available. This reform is particularly difficult to evaluate because crucial data is missing that cannot be collected ex post. It would have been challenging to collect the data before the reform, because the definition of taxable subject (’a data center’) has been changing after 2017.

Furthermore, while a reduced energy tax on electricity is likely to increase profits, the lost tax-revenues might have been partially recoverable with a non-distortive profit tax. However, a foreign center is significantly more likely to be granted an exemption. Because such centers are typically part of an international group, profit taxes are unlikely to be effective as a way to mitigate some of the financial costs of the reform. While our evaluation does not consider employment effects, we note that data centers with no employees have been granted tax reductions.

Overall, the complexity of the reform from a legal perspective has been such that the Tax Authority has been forced to issue clarifications, in turn (essentially) reversing earlier decisions to grant exemptions. We do not explicitly take these changes into account and cannot ascertain how the net benefits of the reform were affected by these changes. Yet, the changing definition of a data center in the energy taxation code, do suggest that the implementation of the reform was rushed. This sentiment also applies to the closing of the exemption in July 2023; the reform was, it seems, closed without considering the Energy Charter, an international agreement that effectively limit governments possibilities to close down energy subsidies.

While the paucity of the data at hand forces us to be very cautious about the net benefits of the reform, we do not find strong evidence that the tax exemption granted to (certain) data centers in 2017-2023 passes a cost-benefit test. The government’s decision to abolish the exemption in 2023 is thus not inconsistent with our findings.