Conclusion
Daylight Saving Time has been advanced as a policy that both saves energy and increases consumer spending. In this brief, the JPMorgan Chase Institute has explored whether or not the latter claim is supported by evidence. We found that card spending in Los Angeles experienced a relative increase of 0.9 percent in the 30 days following the start of DST, and experienced a relative decline of 3.5 percent in the 30 days following the end of DST.
Though we have focused on the comparison to Los Angeles, note that the analysis should be correctly interpreted as one such comparison. We also compared San Diego and Denver to Phoenix to check the robustness of our result. The start and end of DST in San Diego were associated with a relative increase of 2.9 percent and a relative decrease of 2.2 percent in card spending, respectively. In Denver, the relative increase at the start of DST was 0.8 percent, while the relative decrease at the end was 4.9 percent. These comparisons indicate that economic impact of DST is not uniform, and the impact on a given city is an empirical question.
To place these numbers in context, it is useful to consider the impact of a policy that was specifically designed to increase consumer spending. Agarwal and McGranahan (2012) explores sales tax holidays on specific items like clothing and footwear in over 20 states. They find that sales tax holidays boost daily spending on targeted goods by 8 percent, but the holidays last for only three days, muting the long-term effects. Assuming constant spending each day, an 8 percent increase in spending for 3 days translates to a 0.8 percent increase in spending over 30 days. The effect of the start of DST in Los Angeles was to increase daily card spending per capita by 0.9 percent over a 30 day period, an amount that is comparable to the sales tax holiday. Moreover, this effect spans many classes of goods. The effect at the end of DST, however, is a much larger decrease in spending at -3.5 percent. In other words, DST inadvertently has a deeper and broader effect in Los Angeles on spending than a policy specifically designed to stimulate spending.
It is important to note that there are other reasons to support DST beyond the impacts on consumer spending. To the extent that DST increases the amount of daylight available in the evening, it might make it more difficult for criminals to operate. It is plausible that cities that utilize DST may see some benefit from the perspective of public safety. Alternatively, residents may simply enjoy having the additional daylight. In any evaluation of the policy, consumer spending is one among a set of considerations.
Daylight Saving Time is a live issue, as evidenced by the recent debate over a bill (Assembly Bill 385) designed to end the practice in California. To the extent possible, such debates should be driven by empirical evidence. Our analysis suggests that it is far from a given that local commerce will benefit.
Methodology
The JPMorgan Chase Institute conducted the analyses in this brief using anonymized data on credit and debit card transactions in the cities of Los Angeles and Phoenix from October 2012 to June 2015. The sample is restricted to dates that fall within 30 days of the start (early February to early April) or end (early October to early December) of DST each year, leaving a total of four time-change episodes of each kind. In order to focus on customers whose credit and debit card spend is meaningfully large, customers must have at least five transactions in the 30 days before or after a particular time change in a given year to be retained in the sample for that year. After applying these filters, the final sample of data captures the spending behavior of over 450,000 anonymized customers transacting in Los Angeles and Phoenix.
The empirical methodology for estimating the effect of DST is a modified difference-in-differences. To illustrate the approach, the measurement of the effect for the end of DST is as follows. For each of the years in which the end of DST is observed (2012-2015), daily card spending per capita is calculated for each combination of city and time period:
- The 30 days after DST ends in Los Angeles (A);
- The 30 days before DST ends in Los Angeles (B);
- The 30 days after DST ends in Phoenix (C); and,
- The 30 days before DST ends in Phoenix (D).
The end of DST is an annual event that appears four times in our data, and each of the above measures is averaged across all four years. To a first approximation, the modified difference-in-differences estimate is calculated in two stages. The first involves capturing the growth in daily card spending per capita in Los Angeles (A/B - 1) and the growth in daily card spending per capita in Phoenix over the same period (C/D - 1). In the second stage, the ratio of these growth rates is calculated. The result can be thought of as the difference in the growth rates between the “treatment group” (i.e. Los Angeles, which observes DST) and the “control group” (i.e. Phoenix, which does not observe DST). This difference can be interpreted as the causal impact of the end of DST on daily spending.
In the case of the end of DST, this effect is -3.5 percent in Los Angeles, which represents the decline in growth of daily card spending per capita that results from the lost hour of sunlight after DST ends. The same methodology can be employed for the beginning of DST. In that case the effect is 0.9 percent, which represents the increase in growth of daily card spending per capita that results from the extra hour of sunlight after DST begins. This result holds up against robustness checks for “day of week” effects and potentially different growth profiles across the two cities.