State-local sales tax, spillover, and economic activity: examining county governments in the US

ABSTRACT

The reliance on sales taxation has increased in both states and counties. This study aggregated county-level data and empirically explored the associations between state-local sales taxes and economic activity by industry types in county governments for the period of 1990–2013. The results revealed negative associations of sales tax rates with economic activity, especially in the manufacturing industry. Further, the findings provided the spatial dependence of economic activity across counties as a form of possible spillover. This study suggests that policymakers should pay attention to how the manufacturing and retail industries respond to any changes in sales tax rates for business activity.

KEYWORDS:

• Local sales tax
• business activity
• annual payroll
• number of employees
• number of establishments

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1. Since the property tax revolts in the 1970s, sales tax has developed in local governments and local reliance on property tax has substantially declined from 63.61% in 1970 to 34.99% in 2000, and sales taxes make up 39.06% of state-local own-source tax revenue in 2014 (U.S. Census Bureau 2016b). While the number of local governments that levied their own sales tax was around 3,000 in the early 1970s, it increased to over 9,000 after three decades (Luna 2004).

2. The inefficiency depends on a tendency to hold spending below marginal benefits levels because competition results from the scarcity of resource (Oates 1999; Wilson 1999). An inefficiency results from an alteration of taxation of a government (sometimes, unnecessary one) due to its neighbours’ changes, and a taxpayer has different locations where the taxpayer makes an economic choice such as living for property tax, working for income tax, and shopping for sales tax. A government may decrease taxes to enhance more business activity and employment, while a government can increase taxes when the others keep increasing tax rates and it has business favourable tax treatments (Hoyt and Harden 2005).

3. Retailers are a linker between tax-collectors (government) and taxpayers (shoppers). A manufacturing company in a higher tax area pays more for raw materials and the higher sales tax rates result in increases in manufacturing costs. Consequently, sales tax might change their decision on locations and irritate business activity.

4. The border effect among local governments is more dynamic due to the shorter distance of the proximity.

5. The detailed data are an important contribution to the extant literature and examine the discrepancies between state and local sales taxes.

6. The business activity follows the definitions of the Standard Industrial Classification (SIC) and the North American Industry Classification System (NAICS). Before 1998, the definitions followed the SIC, and the classifications for the industries of manufacturing and retail were included in the SIC.

7. The sub-county level is not considered here.

8. This study adjusts the statutory local sales tax rate as ; where is the adjusted local sales tax rate of county in fiscal year . denotes sales tax rate; equals to the days (365 or 366) in a fiscal year , and denotes the days of the new local sales tax rate in effect.

9. A panel data set including multi-span years provides an ideal vehicle to identify the effects of sales tax rate changes, and enables this study to capture spatial effects in addition to recognising the variations over time. The empirical analysis specified time fixed effects to identify the changes in state-local sales tax rates on county business activity.

10. To my knowledge, it has not succeeded to find out a valid instrument that is correlated with sales tax rate but orthogonal to business activity in counties. An alternative empirical approach is to consider dynamic estimates via a system GMM (Roodman 2009). However, the system GMM with the data set in this paper failed to pass overidentification test of Hansen J-statistic, and I could not conclude the internal instrument of the lagged dependent variable is appropriate for the dynamic estimates.

11. The five states that did not impose statewide sales taxes were Alaska, Delaware, Montana, New Hampshire, and Oregon. The 12 states that did not impose local sales taxes were Connecticut, Delaware, Indiana, Kentucky, Maine, Maryland, Massachusetts, Michigan, Montana, New Hampshire, Oregon, and Rhode Island. The four states that did not allow counties to impose local sales taxes were Mississippi, Nebraska, South Dakota, and Vermont. Although Virginia is not included in this list of 21 states, Virginia allows its all localities to adopt local sales taxes, but all the counties and cities are independent from each other. The localities in the Commonwealth of Virginia can impose only 1% on sales taxes and do not have autonomy to change the rate.

12. Unlike state and local sales tax rates in the descriptive statistics, this study obtained the combined sales tax rates after adjusting both the rates by the effective date for regression analysis. Thus, there are some gaps between the sum of state and local sales tax rates and the combined sales tax rate.

13. STATA package – xsmle – can estimate the dynamic effects of a spatial panel models if a panel data set is strongly balanced. An unbalanced panel data set in this study has missing data that result in major problems in the estimation. There is also an econometric strategy to overcome this issue by using multiple imputations that replace missing values by multiple sets of plausible values (Rubin 2004). Since multiple imputations still result in biases in the estimation, this work employed fixed-effect regression analysis with spatial matrix weighting to capture spillover effects across counties.

14. I also analysed the effects of sales tax rates on the business activity in a county with a subsample of counties where state and/or local sales taxes are levied. The findings are consistent to each other, and the coefficient of each variable is similar. Thus, this study provides the only regression results of the full sample in this study, and the subsample results can be provided upon any request.

15. This study actually dropped the entire county if a county had any missing observation over the time period, and this helped construct a strongly balanced panel data set. The investigation analysed the effects by using a spatial autoregressive model – xsmle – in STATA. However, the estimates had the same effects with different coefficients. In order to keep the sample size large, this study conducted a regression analysis using fixed effects after manually creating a spatial lagged variable.

Additional information

Notes on contributors

Jongmin Shon

Jongmin Shon is an assistant professor at the School of Public Affairs and Administration, Rutgers University, Newark, U.S.A. His research interests include public budgeting and finance.