Variables Coefficient Hazard
Treatment 3.314 *** 27.500 0.449
Car 0.512 1.669 0.333
Electronics 0.362 1.437 0.287
Food Store 0.120 1.127 0.250
Personal Care -0.125 0.883 0.294
Clothing 0.027 1.028 0.282
Hobby 0.576 . 1.779 0.302
Miscellaneous Store 0.263 1.300 0.254
Home 0.255 1.290 0.291
Restaurants 0.002 1.002 0.247
Nonstore Retailers 0.599 1.821 0.468
(𝛽0 + 𝛾1)
Car -2.675 ** 0.069 0.900 0.639 1.89
Electronics -2.126 *** 0.119 0.616 1.188 3.28
Food Store -3.175 *** 0.042 0.699 0.139 1.15
Personal Care -18.210 *** 0.000 0.708 -14.896 0.00
Clothing -1.221 * 0.295 0.507 2.093 8.11
Hobby -2.347 *** 0.096 0.606 0.967 2.63
Miscellaneous Store -2.921 *** 0.054 0.672 0.393 1.48
Home -4.032 *** 0.018 1.090 -0.718 0.49
Restaurants -4.669 *** 0.009 1.020 -1.355 0.26
Nonstore Retailers -18.930 *** 0.000 1.182 -15.616 0.00
Survival AnalysisActive Transportation Project Economic Vitality Business Displacement
• How do active transportation projects affect business vulnerability in different industries?
Variable NAICS Industry Name
Car 441 Motor Vehicle and Parts Dealers
447 Gasoline Stations
Home 442 Furniture and Home Furnishings Stores
444 Building Material and Garden Equipment and Supplies Dealers
Electronics 443 Electronics and Appliance Stores
Food Store 445 Food and Beverage Stores
Personal Care 446 Health and Personal Care Stores
812 Personal and Laundry Services
Clothing 448 Clothing and Clothing Accessories Stores
Hobby 451 Sporting Goods, Hobby, Musical Instrument, and Book Stores
Miscellaneous Store 452 General Merchandise Stores
453 Miscellaneous Store Retailers
Nonstore Retailers 454 Nonstore Retailers
Drinking Places 7224 Drinking Places (Alcoholic Beverages)
Restaurants 7225 Restaurants and Other Eating Places
Table 1. Industry Classification
Note1: *** p<0.001, ** p<0.01, * p<0.05, . p<0.1
Note2: According to the result of PH assumption test, yearstart, group are considered using strata.
Note3: Likelihood ratio test = 40.88 on 21 df, p=0.006
Table 2. Result of Extended Cox Model (n=632)
: Industry classification (dummy variable)
Extended Cox Model
• Dependent variables:
- Duration of establishment: amount of time between business start and bankrupt
- Bankrupt: whether the establishment went bankrupt before 2015 (dummy variable)
• Independent variables:
- Treatment: whether the establishment experienced active transportation project (dummy variable)
- Industry: based on 3-digit and up NAICS codes, divided into ten groups (base=Drinking Places)
- Interaction terms: Treatment*Industry
- Year Start: year that the establishment opened
- Group: Central Avenue / Franklin Avenue
• National Establishment Time Series (NETS) database:
- including annually business address, NAICS code, employment and sales by establishment;
- collected from January 1990 to January 2015;
- private, proprietary sources of employment data.
• Franklin Avenue in Minneapolis
- Active transportation project in 2011
- Bike lane installation, removal of parking lane
- Control corridor: another segment of Franklin Avenue
where no active transportation projects were installed
• Central Avenue in Minneapolis
- Active transportation project in 2012
- Bike lane installation, reducing the width of travel lanes
- Control corridor: University Avenue (parallel to Central Avenue)
• Proportional hazards (PH) assumption:
the baseline hazard does not include time-dependent X’s.
• Tr is time-dependent variable:
Even if the establishment is located in the treatment corridor,
it does not experience active transportation project before construction year.
Before active transportation project, the retail store,
After active transportation project, the retail store in the treatment corridor,
• Positive impacts:
- Active transportation projects bring about positive impacts on locally-based retail business.
(Drennen, 2003; NYCDOT, 2013)
• Negative impacts:
- Bike lane intervention only increases the sales of local-serving establishment (Poirier, 2018).
- Proximity to transportation infrastructure could result in an increase in probability of gentrification.
(Grube-Cavers and Patterson, 2015)
Change in hazard ratio
after active transportation projects
Minneapolis City Boundary
Figure 1. Corridor Map
• Results of Extended Cox model:
- Depending on industries types,
there are different effects of
active transportation projects on
vulnerability of businesses.
- Part of the reason is the increase
in inconvenience of car driving
leads to negative impacts on
retails more easily reach by cars.
- Also, it results from the general
shopping atmosphere continuity
in those corridors.
Acknowledgement: The authors would like to acknowledge support from the National Institute for Transportation and
Communities (NITC) under grant number PPMS#1031 and PPMS#1161 and the Summit Foundation. The authors would also
like to thank our partners at PeopleForBikes and Bennett Midland for collaborating on this research effort.
𝑒 𝛽0+𝛾1 𝑡𝑖𝑚𝑒𝑠
• Main finding:
- Vulnerability of businesses to active transportation projects is differentiated depending on industry type.
- Planners should rethink about active transportation projects such as potential displacement effects.
- We emphasize efforts to mitigate the displacement problem of business establishments
as well as promoting local economic growth through active transportation projects.
• Future work:
- Analysis of effects of active transportation projects on changes in business activities by industry types
- Identifying factors that contribute to the differential outcomes for different industry types
Figure 2-1. Number of Establishment Figure 2-2. Employment Figure 2-3. Sales
• Franklin Avenue
- After street improvement, decrease in all economic activity indicators
- However, economic indicators of the control corridor show the similar trends
• Central Avenue
- After street improvement, decrease in all economic activity indicators compared to control corridor