Can Urban Compactness Attract more new Firms?

The Relationship between Degree of Urban Compactness and New Firm Growth in the U.S. Metropolitan Cities
PhD Student
Department of City and Metropolitan Planning
University of Utah
Outcome
Relationship to Compactness
Impact of 10% Score Increase
Average household vehicle ownership
Negative and significant
0.6% decline
Vehicle miles traveled
Negative
7.8% to 9.5% decline
Walking commute mode share
Positive and significant
3.9% increase
Public transit commute mode share
Positive and significant
11.5% increase
Average journey-to-work drive time
Negative and significant
0.5% decline
Traffic crashes per 100,000 population
Positive and significant
0.4% increase
Injury crash rate per 100,000 population
Positive and significant
0.6% increase
Fatal crash rate per 100,000 population
Negative and significant
13.8% decline
Body mass index
Negative and significant
0.4% decline
Obesity
Negative and significant
3.6% decline
Any physical activity
Not significant
0.2% increase
Diagnosed high blood pressure
Negative and significant
1.7% decline
Diagnosed heart disease
Negative and significant
3.2% decline
Diagnosed diabetes
Negative and significant
1.7% decline
Average life expectancy
Positive and significant
0.4% increase
Upward mobility (probability a child born to a
family in the bottom income quintile reaches the
top quintile by age 30)
Positive and significant
4.1% increase
Transportation affordability
Positive and significant
3.5% decrease in transport costs relative to income
Housing affordability
Negative and significant
1.1% increase in housing costs relative to income.
(*) Source: Ewing R., Hamidi S. (2014). Measuring Urban Sprawl and Validating Sprawl Measures.
• Change in spatial patterns of jobs and
firms can influence sprawl(Burchfield et al 2006).
• Four types of development patterns(Ewing 1997)
 Low-density, strip, leapfrog, scattered
 All of them are related to nonresidential development as well as
development patterns(Weitz and Crawford 2012).
• “Firms would decentralized in part to gain
shorter commutes for workers, expecting
to reduce wage costs(Crane and Chapman 2003).”
• However, simple descriptive analysis of
firm locations in terms of changes in
sprawl as a measure of changes in
development patterns is not enough.
vs.
Urban Sprawl
Urban Compactness
•
•
•
•
•
•
•
•
•
•
Many employment centers (polycentric)
Cheap land price available in suburbs
Low density development
Longer commute distance
Few centers (monocentric)
Expensive land price
High density development
Knowledge spillover
Clustering of firms
Shorter commute distance
Locating new firms?
Locating new firms?
Does change in the degree of urban
compactness affect change in new firm growth
over time? If so, which development pattern –
sprawl or compactness – can attract more new
firms and encourage firm growth over time?
• Unit of Analysis: Metropolitan Statistical Area (MSA) in the U.S.
 366 MSAs → 190 MSAs for analysis due to missing values
Dependent Variable
newfirms
% change in the number of new firms (2000 – 2010)
Census Business
Dynamic Statistics
Independent Variable
compactindex
% change in the Ewing’s Compactness Index (2000 - 2010)
Ewing and Hamidi (2014)
Control Variable
exfirm_age1to5
% change in the number of existing firms (age 1-5) in 2000 and 2010
lrgfirms
% change in the number of existing firms (age 1-5) in 2000 and 2010
firmdeath
Change in the firm death ratios between 2000 and 2010
unempden
% change in the unemployment density (2000 – 2010)
patents
% change in the number of patents (all classes) in 2000 and 2010
hhi
Change in Herfindahl-Hirchman Index(HHI) for measuring the degree of industry mix
t2w_under20min
% change in the number of commuters who took less than 20 minutes to work
grossrent
Change in the median gross rent between 2000 and 2010
pop_fborn
% change in the number of foreign-born population
regname
Regional categorical variable (Northeast/Midwest/South/West)
Census
nonres_landprice
Percent change in non-residential (industrial/flex/retail/office) land price: 11 MSAs
Costar 2008-2012
Census Business
Dynamic Statistics
Census Business
Dynamic Statistics
Census Business
Dynamic Statistics
Census 2000
ACS 2010-2014
U.S. Patent and
Trademark Office
Census Business
Dynamic Statistics
Census 2000
ACS 2010-2014
Census 2000
ACS 2010-2014
Census 2000
ACS 2010-2014
Variable
Min.
Max.
Median
Mean
Std. Dev.
% change in the degree of compactness index
-0.7092
0.6041
-0.0304
-0.0280
0.2187
% change in the number of new firms
-0.4901
0.2513
-0.2372
-0.2206
0.1286
% change in the number of existing firms
-0.0925
0.4014
0.1250
0.1184
0.0781
Sprawl and more new firms
Sprawl and less new firms
Compact and more new firms
Compact and less new firms
Percentage Change in the Number of New Firms
(2000 – 2010)
Percentage Change in the Degree of Compactness Index
(2000 – 2010)
Percentage Change in the Number of Existing Firms
(2000 – 2010)
Legends
Negative % change
in compactness
index
Decrease in % of
new firms
Positive % change in
compactness index
Increase in % of new
firms
• The dependent variable ‘newfirms’ shows a quite “normal” distribution
throughout the 190 MSAs
• “The more compact a MSA becomes, the slightly less probability of new
firms being located in that MSA.”
 10% increase in the compactness index → 0.8% decrease in the number of new firms
(Intercept)
Compactness Variables
% change in compactness index
Existing Firms Characteristics variables
% change in the number of existing young firms (age 1-5)
% change in the number of existing large-employee-size firms (employee 250+)
change in firm death ratios
Economic variables
% change in unemployment density
% change in the number of patents
change in Herfindahl-Hirchman Index (HHI)
Accessibility to workplace variable
% change in the number of workers with average-travel-to-work under 20 minutes
Demographic/Housing variables
change in the median gross rents
% change in the number of foreign-born population
N
R-squared (R2)
Std. coeff.
-0.13749
Std. error
0.026033
t-value
-5.281
p-Value
3.69E-07 ***
-0.08142
0.026572
-3.064
0.00252 **
0.705904
-0.04674
-1.09008
0.051182
0.037704
0.606985
13.792
-1.24
-1.796
< 2e-16 ***
0.21672
0.0742
-0.4465
0.00826
0.001219
0.457976
0.010583
0.000247
-0.975
0.78
4.942
0.33091
0.43613
1.76E-06 ***
0.023631
0.05062
0.467
0.64118
0.001037
0.033712
0.000549
0.018427
1.89
1.829
0.06034
0.06899
190
0.681
• Including the categorical variable of the 4 regions in the U.S.
 The degree of compactness is still significant.
 The regional effect exists.
(Intercept)
Compactness Variables
% change in compactness index
Existing Firms Characteristics variables
% change in the number of existing young firms (age 1-5)
% change in the number of existing large-employee-size firms (employee 250+)
change in firm death ratios
Economic variables
% change in unemployment density
% change in the number of patents
change in Herfindahl-Hirchman Index (HHI)
Accessibility to workplace variable
% change in the number of workers with average-travel-to-work under 20 minutes
Demographic/Housing variables
change in the median gross rents
% change in the number of foreign-born population
Regional categorical variable (Reference: Midwest Region)
Northeast region
South region
West region
N
Adjusted R-square(R2)
Std. coeff
-0.11317
Std. error
0.028906
t-value
-3.915
-0.05803
0.028518
-2.035
0.764223
-0.06337
-0.7887
0.053118
0.036601
0.594663
-0.30572
0.012389
0.001033
0.443693
0.010224
0.000242
-0.689
1.212
4.269
0.065449
0.050343
1.3
0.000952
-0.00335
0.000548
0.020477
1.739
-0.164
0.083848 .
0.870227
0.035132
0.016785
-0.04552
0.018886
0.01644
0.019919
1.86
1.021
-2.285
0.064517 .
0.308678
0.023481 *
190
0.689
p-value
0.000129 ***
0.043367 *
***
14.387 < 2e-16
-1.731 0.085125 .
-1.326 0.186461
0.491706
0.227213
3.21E-05 ***
0.195282
• Identifying the NAICS sectors of firms affected by change in the degree of
compactness
 “Utilities (NAICS 22)”, “Health Care and Social Services (NAICS 62)”, “Accommodation
and Food Services (NAICS 72)” show significant relationship with change in the degree
of compactness.
New Firms: NAICS 22
Std. coeff
(Intercept)
compactness index
existing young firms (age 1-5)
New Firms: NAICS 62
Std. coeff
New Firms: NAICS 72
Std. Error
t-value
p-value
Std. Error
t-value
p-value
Std. Error
t-value
p-value
-0.103856 0.0912896
-1.138
0.25678
0.0390522 0.0200893
1.944
0.05347
0.0377543 0.0212428
Std. coeff
1.777
0.07722
0.2791478 0.0931788
2.996
0.00313 **
-0.050835 0.0205051
-2.479
0.01409 *
-0.072457 0.0216824
-3.342 0.001013 **
-0.098973 0.1794815
4.57E-08 *** 0.1160021 0.0417647
2.778 0.006061 **
0.03053 *
0.0556242 0.0307667
1.808
0.072295
-0.812
0.417919
0.825
0.410563
-0.551
0.58202
0.2256112 0.039497
5.712
existing firms with large employee size
(employee 250+)
0.1458701
0.132218
1.103
0.2714
0.0634433 0.0290961
2.18
firm death ratios
0.9040063 2.1285232
0.425
0.67156
-0.143094 0.4684067
-0.305
0.76035
-0.402141 0.4952999
median gross rents
-0.000513 0.0019243
-0.267
0.78996
-0.000612 0.0004235
-1.444
0.15035
0.0003693 0.0004478
unemployment density
5.0958707 1.605993
3.173
0.00178 **
0.2178484 0.3534177
0.616
0.53841
-0.978888 0.373709
Herfindahl-Hirchman Index (HHI) of industrial mix
0.0009023
0.000865
1.043
0.29828
0.0001996 0.0001904
1.049
0.29567
0.000358 0.0002013
patents
-0.011647 0.0371126
-0.314
0.75402
0.0088721 0.0081671
1.086
0.2788
0.0214476 0.008636
2.484 0.013929 *
workers with travel time to work under 20 min
-0.076452 0.1775115
-0.431
0.66721
0.1092306 0.0390635
2.796
0.00574 **
0.0160281 0.0413063
0.388
foreign-born population
-0.169407 0.0646184
-2.622
-1.143
0.25472
0.0579099 0.0150365
3.851 0.000163 ***
N
2
R-squared (R )
0.0095 **
-0.016248
0.01422
190
190
190
0.148
0.425
0.347
-2.619 0.009565 **
1.779
0.077016
0.698455
• Sprawling MSAs tend to attract new firms with a small number of
employees (1 – 4 employees).
• Compact MSAs have a benefit of attracting more new firms with larger
employee size(20 – 49 employees) than sprawling cities.
New firms (1 - 4 employees)
Std. coeff
Std. Error
t-value
New firms (20 - 49 employees)
p-value
Std. Error
t-value
p-value
-0.14777 0.147469
-1.002
0.31769
(Intercept)
-0.13829 0.027912
-4.955
compactness index
-0.06271 0.029429
-2.131
0.808837 0.052618
15.372
-0.04871 0.036968
-1.318
0.1893
-0.03253 0.198729
-0.164 0.870163
-0.6821 0.623798
-1.093
0.2757
-1.37489 3.271644
-0.42 0.674815
0.122
0.9029
0.750652 2.417324
0.311 0.756521
0.002846 0.001297
2.194 0.029526 *
existing young firms (age 1-5)
existing firms with large employee size (employee 250+)
firm death ratios
unemployment density
Herfindahl-Hirchman Index (HHI) of industrial mix
0.056073
0.45883
1.67E-06
Std. coeff
***
0.0345 *
0.302653 0.154342
1.961 0.051448
< 2e-16 ***
0.962477 0.276033
3.487 0.000616 ***
***
0.001189 0.000247
4.814
3.13E-06
-0.00667 0.019965
-0.334
0.7388
-0.11935
0.10521
-1.134
Northeast
0.038365 0.019709
1.947
0.0531 .
0.062614
0.10339
0.606 0.545548
South
0.019273 0.016717
1.153
0.2505
0.226659 0.087687
-0.023 0.020568
-1.118
0.265
-0.12974 0.107982
foreign-born population
0.25814
Regions (Reference: Midwest)
West
N
2
R-squared (R )
190
190
0.683
0.173
2.585 0.010543 *
-1.201 0.231162
•
•
Generally, urban sprawl supports new firm
growth. However, in terms of employee size,
the model shows that compact MSAs
support more new firms with a larger
employee size (20-49 employees) than
sprawling cities.
In the overall regression model, two control
variables are significant
 Positive change in the number of
existing young firms (age 1 – 5):
positive change in the number of new
firms
 The better HHI index, the more new
firms can attract into MSAs.
•
•
•
Regional effect exists, but not all regions
are significant.
Unlike the existing studies, this study shows
that some variables (ex. foreign-born
population, the number of patents,
unemployment density, and the number of
patents) are not significant in the models of
this study.
Generally, non-residential land price in
compact MSAs tend to have the lower
number of firms in 2010 compared to
sprawling cities, but due to outliers,
interpretation of this result is still debatable.
• Using two time periods
 Not sufficient to understand the causal relationship between the degree of
compactness and new firms growth over time
 Need the most recent sprawl index
• Using a simple OLS regression
 All MSAs belong to one of the four regions or the nine divisions defined by Census
Bureau.
 Calculating the group random effects of the region (or divisions) to measure its
magnitude and variation by using the multilevel modeling.
• Adding more variables that affects changes in the number of new firms
 Although the compactness index includes a centering factor, identifying the number of
employment centers and the CBD of each MSAs can show the direct relationship
between urban pattern and new firm growth over time.
 Additional variables (ex. Percentage change in non-residential (industrial/office/flex/
office) land prices between 2000 and 2010) that are related to the degree of
compactness index should be needed.
Keuntae Kim
PhD Student
Dept. of City & Metro. Planning
University of Utah
[email protected]