Local excise taxes, sticky prices, and spillovers: evidence from Berkeley’s soda tax

An income effect will also reduce all consumption. It is trivial in the present context.

See also (National Research Council (NRC), 1981; Young & Bielińska-Kwapisz, 2002; Kenkel, 2005; Barzel, 1976; Sumner & Ward, 1981; Keeler et al., 1996; Delipalla & O’Donnell, 2001).

https://​www.​nytimes.​com/​2014/​10/​18/​us/​michael-bloomberg-to-back-soda-tax-in-berkeley.​html.

The significance of these results are not robust across alternative research designs.

The Philadelphia tax was approved by the Philadelphia City Council in June 2016 and levied a $0.015 tax on distributors, beginning in January 2017. Unlike the Berkeley tax, the Philadelphia tax included artificially sweetened beverages in addition to sugar-sweetened beverages.

Anonymized store data are geolocated only by 3-digit zip code. Berkeley’s municipal boundary is nearly contiguous with the 3-digit zip code boundary, but not perfectly so. Some parts of neighboring Albany and Kensington are located within the same 3-digit zip code as Berkeley. By examining counts of brand outlets in various 3-digit zip codes across the U.S., we are able to infer the identity of retailers. That knowledge, combined with knowledge about products carried at the retail stores in question allow us to identify one supermarket and one mass-merchandise store located within the 3-digit zip code encompassing Berkeley that operate outside the Berkeley city limits. Hence, these stores are not subject to the tax and, therefore, are not included among our treatment group.

For matching, we smooth the price and volume time-series with a four-week running average.

In practice, all comparison stores are drawn from the same retailer, so retailer-UPC-week effects collapse to UPC-specific time fixed effects.

Though assessment of the tax was ultimately delayed until March 2015, the Berkeley supermarket increased prices dramatically and very briefly in January.

Results are very similar when using the store-level matches, which uses the same control stores for all UPC regressions.

These series reflect the group mean of standardized, four-week-moving-average, volume-weighted prices, and the group mean of standardized, four-week-moving-average consumption, respectively.

University of California student matriculation may cause seasonal deviation relative to potential comparison stores.

We get similar results with UPC-level analysis using the original sets of controls that are common to all UPCs at a given retailer.

Store brands are much more relevant in soda than the other categories studied by Butters et al. (2022): beer, liquor, and cigarettes.

Price and consumption responses are greatest immediately following tax imposition, and decline in magnitude by at least half by the end of 2015.

Lovenheim (2008) estimated that between 13% and 25% of cigarette consumers residing near a state boundary with a lower tax state make purchases across borders, evidencing consumer search to avoid high prices. And in an evaluation of cigarette and beer excise taxes, Harding et al. (2012) observed Nielsen “Homescan” households crossing state borders in order to avoid relatively steep excise taxes in their home states. Tax pass-through is estimated to increase with retailer distance from low-tax borders, reflecting cross-border competition among consumers with relatively low costs of tax-avoidance. Cigarette tax pass-through is estimated to be 49% at retailers near boundaries with low tax states, increasing 7.6 percentage points for every one-percent increase in distance from the state border. A similar, but more pronounced effect is observed for beer. Similarly, Dutkowsky & Sullivan (2017) observe under-shifting of cigarette taxes among retailers in proximity to low-tax states, and DeCicca et al. (2013a) show cigarette taxes are shifted at lower rates to consumers who make cigarette purchases in states other than those in which they reside.

Harding et al. (2012) identified stockpiling behavior among intensive cigarette consumers, who purchased relatively greater quantities of cigarettes when shopping in neighboring jurisdictions. Chiou et al. (2014) also find evidence of cigarette stockpiling in advance of an anticipated excise tax change in Chicago. Stockpiling is less prevalent in proximity to low-tax jurisdictions, where its benefits are presumably lower because post-tax avoidance is less costly.

See http://​www.​bayareacensus.​ca.​gov/​cities/​Berkeley.​htm

If we restrict our attention to consumers who frequent (at least once in 2015) the Berkeley chains we observe in the Nielsen retail panel, the shares of cigarette and soda consumers are similar.

This analysis requires data forensics because of measures taken to guarantee retailer anonymity in the Nielsen-Kilts data upon which our analysis relies. Stores are identified only by an anonymous store identification number and geolocated only by a three-digit zip code. Retail chains are identified by an anonymous retailer code. Though the three-digit zip code that encompasses Berkeley shares some boundaries with the city, the zip code region also incorporates small parts of neighboring Albany, Kensington and Oakland. We identify retailers by analyzing store counts across many, varied three-digit zip codes and comparing these counts to Google search results and retailers’ store locator services. We further compare sales of products observed in our data that are known to us to be sold at only one of the supermarkets. In this way, we can distinguish a treated supermarket in Berkeley from a nearby untreated supermarket in an adjacent city, and, therefore, to compare consumption changes due to the tax.

Our analysis of Nielsen “Homescan” data for California indicate that 22% of soda volume is purchased from drugstores.