NATIONAL STREET IMPROVEMENT STUDY WW Ill! - -. 5.241; 1 1 i1 - rr 4?12%} . 25?1-1- 1 - ?7 .- if DATA SOURCES OVERVIEW1 Across the country, policymakers and planning departments are making cities more livable by better accommodating people who walk and bike. Improving streets and upgrading transportation infrastructure often require reducing onstreet parking or traffic lanes. While studies have shown how such upgrades improve traffic safety and mobility for city residents, the question remains how such infrastructure improvements affect economic outcomes. This study will attempt to answer to what extent these types of corridor-level street improvements impact economic activity and business vitality in the immediate vicinity. In particular, how do street improvements impact retail sales and employment? Memphis has conducted many street improvement projects in past years, including new protected or buffered bike lanes. This report explores two recent street improvement corridors—Madison Avenue and Broad Avenue—to understand the economic and business impact of these active transportation infrastructure investments. Assessing the impact of street improvements and the accompanying reduction of on-street parking or travel lanes on a neighborhood’s economic activity and vibrancy is a new field of research. In 2013, the New York City Department of Transportation commissioned a first-of-itskind study, using sales tax data to evaluate how businesses on improved corridors have been affected. This current study builds on past work by examining additional cities and incorporating new research methods and data sources. KEY FINDINGS Based on our analysis, we found the street improvement projects in Memphis did not impede economic vitality, and may have contributed to positive growth. In particular, we can conclude that: 1 2 For this study we used multiple data sources to estimate the effect of new bike lane infrastructure investment, each with its pros and cons.3 As such, the analysis results using the three data sources should be viewed as complementary to each other. BOTH TREND ANALYSIS AND INTERRUPTED TIME SERIES ANALYSIS REVEALED CONSISTENT EVIDENCE OF POSITIVE IMPACTS OF THE PROTECTED BIKE LAND ON BROAD AVENUE ON EMPLOYMENT IN THE FOOD SERVICES INDUSTRY, INDICATING AN IMPROVEMENT IN BUSINESS VITALITY AS A RESULT. WHILE SOME ANALYSIS METHODS SHOWED SIMILAR TRENDS IN EMPLOYMENT AND SALES BETWEEN MADISON AVENUE AND ITS CORRESPONDING CONTROL CORRIDORS, WE FOUND CAUSAL EVIDENCE THAT THE BUFFERED BIKE LANE INSTALLED ON MADISON AVENUE TRIGGERED SIGNIFICANT INCREASES IN BOTH FOOD SERVICES EMPLOYMENT AND RETAIL SALES ACROSS MULTIPLE DATA SOURCES AND ANALYTICAL METHODS. First, we used the Longitudinal OriginDestination Employment Statistics (LODES) data set from the Longitudinal Employer-Household Dynamics Dataset (LEHD). LEHD provides geographically granular detail about jobs, workers and local economies, allowing us to examine employment by broad industry sector, wage and educational attainment. One major disadvantage of the LEHD data set is that in order to guarantee confidentiality, block level data is “fuzzed” so the numbers SALES TAX DATA In the other analyzed corridors and industry sectors, we found either mixed or insignificant results.2 However, the insignificant results may be significant in this context. Importantly, there is no evidence of a negative economic impact from right-of-way or parking lane removal. 1. The National Street Improvements Study is a research project by Portland State University, Bennett Midland, and PeopleForBikes. An accompanying report with more detailed information on methodology can be accessed at https://peopleforbikes.org/placesforbikes/ resources/ 2. This is typically due to control corridors that may not be fully comparable (for DID analysis), methods explained further in Section 3 (“Methodology”). Memphis Corridors LEHD 3. Because this project makes use of a variety of different data sources, it required collaboration between the research team and representatives from multiple agencies/departments. Our principle contact was with the City of Memphis, sales tax data was provided and aggregated by State of Tennessee Department of Revenue; QCEW data was provided by Tennessee Department of Labor and Workforce Development; and LEHD data was publicly available at United State Census Bureau. 4. For the North American Industry Classification System (NAICS), please visit https://www.census.gov/eos/www/naics/. Sales tax data is collected as the primary data source to allow us to estimate a more sensitive measure of economic activity than employment (as the decision to hire or fire employees for a firm is often an expensive one, and thus we would expect employment to be a delayed response to changes in economic activities). Some drawbacks of sales tax data are that some states do not have a sales tax or, in states or cities that do have one, the sales tax data is not broken down by specific industry and it is difficult to parse out accurate figures. But the benefits of sales tax data largely outweigh these issues and do offer a more sensitive metric than employment. Tennessee has a general 7% sales tax for all businesses, with an QCEW This report also takes advantage of establishment level Quarterly Census of Employment and Wages (QCEW) data. The QCEW gives us address level-data on individual establishments as well as detailed employment information, allowing for more accurate pinpointing of the geographic location of businesses and industrial classifications. In addition, the research team is able to use employment and wages as additional indicators of economic performance in the corridors. We obtained establishment-level QCEW data from the Tennessee State Department of Labor & Workforce Development, and were able to aggregate the employment and wages to the corridor block level do not reflect the exact number of jobs at this geographical level. Additionally, though employment is disaggregated by industry, it is only provided at the most general level (the equivalent of two digit NAICS4 codes) so we are unable to isolate restaurant workers from hotel service workers, for example. That being said, the LEHD data set is comprehensive, offers unprecedented geographic detail, and is longitudinal, allowing for consistent comparisons over time. exception on non-restaurant food which is taxed at 5.5%. In Shelby County, where Memphis is located, there is an additional sales tax of 2.25%, as well as an additional 5% accommodations tax. These produce a sales tax range of 7.75-14.25% in the city of Memphis. However, medical supplies, certain groceries, and food items are exempted from tax collection, which may hamper the ability of sales tax data to accurately reflect all retail business vitality. and block-facing level for retail and food services industries. The aggregated employment numbers correspond closely to the LEHD data used in our analysis, but with the advantage that the numbers are not “fuzzed” for confidentiality concerns. METHODOLOGY Three analytical methods were applied in order to isolate the impact of street improvements while controlling for other economic and regional factors. The methods are an aggregated trend analysis (following the 2013 NYC Department of Transportation study), a difference-in-difference approach, and an interrupted time series analysis. The time frame used in the analysis for LEHD data is 2004-2015, the period is 2004-2016 for sales data, and 2000-2017 for QCEW data. In order to properly isolate the effect of the street improvements we must identify treatment corridors (corridors that actually were improved) and control corridors (corridors that are similar to the treatment corridors except they remain unimproved). Treatment corridors are corridors where new bike or pedestrian related improvements were installed, ideally made up of a minimum of 10 adjacent, or intersecting, census blocks with a minimal number of retail and food service jobs. Additionally, we chose street improvement corridors installed between 2008 and 2013 in order to guarantee we have sufficient data (at least 3 data points pre- and post-treatment) to track pre- and posttreatment economic trends. Once corridors are selected based on these criteria, further testing is conducted to discern the level of similarity between treatment and control corridors. The tests include quintile comparisons of corridor-level employment to city-wide employment, and statistical tests of average block level employment that compare control corridors to the treatment corridors. AGGREGATED TREND ANALYSIS This first analytical method, aggregated trend analysis, follows a previous NYC Department of Transportation study (NYCDOT, 2013), examining whether the treatment corridors tend to have better business performance than comparison corridors after street improvements. The approach compares the trends of treatment and control corridors in addition to city-wide trends over the full time period covered by the data. If treatment corridors show greater increases in employment or sales tax receipts, then that would represent a positive impact of street improvement on business activities. This method is easy to follow and represents the aggregated trend of business activities. However, it lacks the rigor of econometric estimates and statistical tests that explicitly test whether the street improvement caused the change. DIFFERENCE-IN-DIFFERENCE APPROACH The second method aims to estimate the difference in business vitality of pre- and post-improvement periods between treatment and control corridors within the same time period. This is known as a difference-in-difference (DID) approach. The approach looks at the change in the variable of interest—employment levels or sales revenues in our case—in the treatment corridor before and after the street improvement. Meanwhile, the control group has not been treated in either time period. The difference in growth trajectories between the two periods should provide us with an unbiased estimate of the effect of the street improvement. INTERRUPTED TIME SERIES The third method, interrupted time series (ITS), is an econometric technique that estimates how street improvements impact corridor economic vitality from a longitudinal perspective. This approach treats the street improvement as the “interruption” and estimates the change in the level and the growth trend of business activities in the corridor after the street improvement. If the street improvement treatment has a causal impact, the postintervention sales revenue or employment should show a different level or slope compared to the pre-intervention data.5 5. The aggregated trend analysis is a visual and growth trend comparison approach where statistical significance cannot be assigned. However, for the two econometric approaches, DID and ITS analysis, we refer to statistically significant impacts whenever positive or negative impacts are stated in this report. CONCLUSION In conclusion, aggregated trend analysis and DID analysis both utilize control corridors to determine the impacts of the street improvement corridor, while the ITS analysis uses multiple time points on the street improvement corridor itself to pinpoint economic outcomes. In general, the ITS analysis provides more robust results than the other two methods, since it is less likely to be affected by the selection of control corridors. However, this method generally requires more data points post-intervention to achieve meaningful and valid impact estimations. The DID approach is heavily dependent on finding comparable control corridors (which may not always exist), so the analytical results may be weakened when appropriate corridors cannot be identified. Additional data points after the completion of street improvements may help to provide further validity and rigor to the analysis of resulting economic outcomes. Moreover, further contextual information about the street improvement corridor, such as quality or level of the improvement, number of parking spots eliminated, and subsequent bicycle ridership or pedestrian increases, would help to better understand the linkages between the improvements and potential economic impacts. Extending this research to more closely examine the changes and shifts in industrial patterns will be valuable as well. BROAD AVENUE A protected bike lane was installed in 2010 on Broad Avenue—a relatively short corridor, covering five blocks. The control corridors, Cooper Street and Central Avenue, are generally further away from the treatment corridor and have higher traffic volume. CORRIDOR ANALYSIS KEY TAKEAWAYS »iThe pproach analysis shows reveals a positive consistent growth evidence trend im contributes to inaccuracies at the small pproach of positive shows impacts a positive of thegrowth protected trendbike im geographic scale of the Broad Avenue pproach lane street shows improvement a positive growth on employment trend im corridor. pproach in the food shows services a positive industry, growth using trend both im » In conclusion, onofCentral Avenue, we foun The low number retail establishments pproach trend analysis shows aand positive ITS methods growth trend across im conclusion, o n, on Central Avenue, foun along Broad Avenue in the baselineweperiod pproach both employment shows a positive data sources. growth trend data. conclusion, n, on Central Avenue, we foun means thato the large post-construction conclusion, o n, on Central Avenue, we foun » In The terms Broad of Avenue sales data, corridor the aggregated shows growth rates should be interpreted with construction, an improvement trend some anal contradictory terms of sales patterns data, where the caution, andindicating with particular attention to in the business vitality. aggregated trend anal ysis er er the sales tax receipts andapproacer QCEW retail local context. eremployment er er er er er grew er er after er er the er erprotected er er er er er » In conclusion, on Avenue, we theCentral protected bike lane erbike er erlane er erinstallation er er er er but er erthe er LEHD er er erretail er, found a significant positive impact on triggered a significant employment additio . However, additio nal employment decreased. Theeconometric sales tax restaurant Central Avenue following increase sales in theon food services industry after approaches impact is more not st and QCEWsuggest data arethe likely much bike lane construction, indicating an installation, indicating an improvement in reliable in this circumstance, as it is improvement in business vitality. business vitality as a result. quite likely that the fuzzed LEHD data There is a very apparent trend that restaurant sales on Central Avenue increased dra MADISON AVENUE REFERENCES7 NYCDOT. (2013). The Economic Benefits of Sustainable Streets. New York City Department of Transportation. Retrieved from http://www.nyc.gov/html/dot/downloads/pdf/dot-economic-benefits-ofsustainable-streets.pdf Madison Avenue, located in the Midtown district, received a buffered bike lane in 2011. The control corridors are Union Avenue and Cooper Street, close to the treatment corridor in Midtown, and Highland Street, located to the southeast of the district.6 Angrist, J., & Pischke, J.-S. (2009). Mostly Harmless Econometrics: An Empiricist’s Comparison. Princeton University Press. Lopez Bernal, J., Cummins, S., & Gasparrini, A. (2016). Interrupted time series regression for the evaluation of public health interventions: a tutorial. International Journal of Epidemiology, dyw098. https://doi.org/10.1093/ije/dyw098 KEY TAKEAWAYS » Our ITS analyses show positive and statistically significant impacts of the street improvement on both food employment and retail sales, suggesting a positive causal relationship. » Analysis of sales tax receipts also shows that sales along Madison Avenue follows a parallel trend when compared with its control corridors, with no detrimental impacts to either retail or food services industry sales after the street improvement. » The LEHD and QCEW employment data on Madison Avenue both show similar trends with its control corridors after street improvement. Cooper Street experienced a larger bump in employment in the post-construction period, 6. Initially considered as a candidate for a control corridor, Jackson Avenue was ultimately eliminated because of too many dissimilarities with the treatment corridor, such as smaller amount of retail and food employment and geographical distance. but we suspect that this may be due to events unrelated to the construction of the buffered bike lane on Madison Avenue. » DID analyses indicate non-significant or mixed impacts of the Madison Avenue street improvement on the employment and sales economic indicators. » While some analyses indicate little impact of the street improvement on certain economic indicators, the positive causal results of our ITS analyses are significant enough for us to conclude that the buffered bike lane on Madison Avenue improved food services employment and retail sales in the corridor and had a positive effect on business vitality. 7. A complete reference list is available as part of the accompanying report at https://peopleforbikes.org/placesforbikes/resources/ SUMMIT ATION