Test AI Events Search

Abstract:
How do consumers respond to extreme changes in energy prices, such as those observed during the European energy crisis in 2022 or during other emergency situations?

In the Republic of Georgia, during the COVID-19 lockdowns from November 2020 to February 2021, the government paid the natural gas bills directly to the utilities on behalf of the residential customers, as long as consumption was less than or exactly 200 m3/month, effectively making natural gas free—a 100% cut in price. If consumption was more than 200 m3/month, the consumer had to pay the usual price per unit on their entire consumption (not just the amount in excess of 200). The regular pricing scheme (a one-part tariff with a constant price per m3) was resumed in March 2021.

This setting is interesting for two reasons. First, this temporary policy created the one and only tariff revision that consumers had ever experienced since at least 2014. Second, the policy created a discontinuity—a “notch”—in the consumer’s budget constraint at 200 m3. Economic theory suggests that consumer should “bunch” at 200 m3 (Kleven and Waseem, 2013), but in practice, consumers often find it difficult to grasp block-rate tariffs or other non-linear energy pricing schemes (Ito, 2014; Shaffer, 2020).

Histograms of consumption before, during and after the policy show no bunching before the policy takes place, and pronounced bunching at and just below 200 m3 while the policy is in place, which promptly disappears when it is lifted.

To make sure that such adjustments in consumption are due to the temporary gas pricing policy and not to something else, we devise a difference-in-difference causal inference design, where the treatment is the temporary policy, the control group are consumers living at a small number of locations that are exempt from the policy, and the treatment group is comprised of consumers to whom the policy applies (and living at locations with similar climate and population density as the exempt locations). This approach implies a price elasticity of demand of -0.06 to -0.11.

These estimates are compared with those from three alternative methods—the “notching estimator” developed by Kleven and Waseem (2013), maximum likelihood, and histogram matching—under different assumptions about the heterogeneity of the price elasticity and/or the presence of a significant share of consumers whose elasticity is zero.

Overall, the price elasticity is very modest even as the histograms show a strong degree of bunching at the notch, but is consistent with what the government must have assumed it to be as it sought to establish the threshold that would maximize the consumers’ utility given a government budget constraint. We find no evidence of substitution into other fuels, which suggests that people were monitoring their gas meters carefully in hopes of making or staying below the 200 m3 mark.