by mark d. partridge the ohio state university co-authors:
DESCRIPTION
Dwindling U.S. Internal Migration: Evidence of a Spatial Equilibrium? prepared for presentation at the Gosnell Lecture Series Rochester Institute of Technology Department of Economics Rochester, NY 27 October, 2011. by Mark D. Partridge The Ohio State University Co-authors: - PowerPoint PPT PresentationTRANSCRIPT
1
Dwindling U.S. Internal Migration: Evidence of a Spatial Equilibrium?
prepared for presentation at theGosnell Lecture Series
Rochester Institute of TechnologyDepartment of Economics
Rochester, NY27 October, 2011
by
Mark D. Partridge The Ohio State University
Co-authors:Dan S. Rickman, Oklahoma State UniversityM. Rose Olfert, University of Saskatchewan
Kamar Ali, University of Lethbridge.
2
Introduction
• Long-standing high internal migration rates in US– Such migration reallocates labor to areas with high
productivity. • The pre 20th settlement of the U.S. and the industrialization pre WW II are
the most vivid examples.• Rural-urban migration related to farm/manufacturing realignment.
– Smoothes out regional shocks to labor demand (e.g., from differences in industry composition).
– Such ‘disequilibrium’ migration is often hailed as why US labor markets are flexible in responding to economic shocks compared to Europe (Obstfeld and Peri, 1998).
3
Intro—cont.• Demands shocks should be random, but US migration is
very persistent (Blanchard and Katz, 1992; Partridge and Rickman, 2003).
• The US also has a long history of ‘equilibrium’ migration that originates on the supply or household side—climate and landscape being the largest factors (Graves) – Relates to major debate on jobs vs. people (Greenwood et al., 1991)– e.g., amenity demand related to higher income & pop. aging,
technological innovations such as air conditioning (Rappaport, 2007; Chen and Rosenthal JUE (2008).
– Rappaport (2004) shows how equilibrium migration can persist for decades.
– Could be man-made amenities—e.g., Florida, 2002; Glaeser et al., 2001; and Adamson et al., 2004.
4
Introduction—cont.• Modeling takes place in a spatial equilibrium context. In
equilibrium, utility and profits are equalized across space.– Migration flows are a positive function of utility differentials
• The question is why does the U.S. has such persistent flows to ‘nice’ places? Would not forward-looking agents and markets eventually capitalize the effects of nice climate, income changes, and demographic shifts in wages & housing prices (Evans 1990)?– The US has experienced a long-term downward trend
in internal migration – Began in late 1980s and really accelerated post-2000. We do not
believe this is a housing bubble phenomenon.
5
Annual Gross Migration Rates, U.S. 1947-2008
1.0
2.0
3.0
4.0
5.0
6.0
7.0
Percent
Moved to diff county Moved to diff county diff state
Source: U.S. Census Bureau, Current Population Survey.
6
19911992
19931994
19951996
19971998
19992001
20022003
20042005
20062007
20082009
0
0.002
0.004
0.006
0.008
0.01
0.012
Net Migration (Standard Deviation)
Figure 2: Standard Deviation of Annual State Net Migration as a Share of Initial Population
Source: U.S. Census Bureau Intercensal Population Estimates (http://www.census.gov/popest/estimates.html)
7
Intro—cont.
• Does diminished migration imply the U.S. has achieved a spatial equilibrium of roughly equal profits and utility across space?– There is still economic migration in the ‘flexible’ U.S
labor market, but site-specific factors such as amenities or distance to cities would have a much smaller influence—i.e., the decline of equilibrium migration with capitalization of their effects into wages and housing.
8
Intro Continued
• An alternative explanation is that there has been a decline in economic migration (only a slight decline in the standard deviation of economic shocks). – Superficially, the US labor market would be less flexible.– Take on a European flavor of less geographic mobility with
local labor demand shocks being satisfied by changes in local labor force participation and unemployment rates.
• Implications for policy interventions: place-based vs. people-based
9
Theoretical framework• In spatial equilibrium, indirect utility of residence (V)
should be equal to some level V across locations k (Partridge and Rickman, 1997; Glaeser 2007):
• V = V(.)k-Mk for all k,• V is a function of wages, housing costs, and quality of life. • The overall net migration rate from location i to j (Net
Migij) during the adjustment in any time period depends on the differences in V over all time pds:
• Net Migij(t) = f(∑(Vj-Vi - Mij), – Net Mig is a positive function of utility differentials (Douglas,
1997)
10
Theoretical framework• In equilibrium, firm profits are also equalized across space
for every region j. • Π = ΠJ
• Like movement of households, net movement of firms and establishments from region i to j are a positive function of the respective profit differential
• NET FIRMij = g(Πj- Πi)• We assume this is roughly equalized over the long-run with
the exception of responses to economic demand shocks.• Bringing labor demand and supply together:Migkt = f(LDSHOCKk0, INDUSTRYko, AMENITYk0, URBANk0, DEMOGk0, .),
11
Empirical ImplementationBase Model Population Growth at the county level:Population growth is measured more accurately than net migration and
it includes immigrants who also would be affected by the spatial equilibrium.
α, φ, θ, ψ, and γ are coefficient vectors
Explanatory variables are generally measured in the initial period. Most key variables should be predetermined such as natural amenities.
σs represents the state fixed effects; so coefficient vectors reflect average county variation within states;
ε - the residual, assumed to be spatially correlated within BEA areas but not between them.
isi0i0i0i00)-i(t DEMOG ECON AMENITY GEOG Pop%
12
Explanatory Variables
•DEMAND SHOCK/ECON: using total employment growth would be endogenous.
•Most studies on say ‘jobs versus people’ would need to instrument for total employment using industry mix. In our case, we are not interested in the effects of employment growth per se, but the impact of a demand shock. Industry mix is a demand shock.
•We use Industry mix employment growth, calculated by multiplying each industry's national job growth (1990-2000 and 2000-2007) by the initial period industry shares in each county (Bartik, 1991; Blanchard and Katz, 1992; Bound and Holzer, 2000);
•∑i(ei/E)*gni,
•ei represents county employment in industry i, E is total county employment, and gni is the national growth rate in industry i.
13
Explanatory Variables• AMENITIES: a natural amenity scale constructed by U.S.
Department of Agriculture ERS: 1 = Low level of amenities, 4 = above average, 7 = High;
• proximity to Great Lakes, Oceans• GEOG variables: Distance to nearest Urban Center, Incremental
dist. to an MA, Incr. dist. MA> 250K, Incr. dist. to MA> 500K, Incr. dist. to MA>1.5m (representing distances to urban hierarchy)
• population of actual, surrounding counties & county area • DEMOG incl. 5 ethnicity vars, 4 education vars, % female, %
married, and % with a work disability, all measured in 1990 (2000).– Migration specialists have long pointed to the importance of
demographics—human capital and life cycle. However sorting of the young, in particular, causes us to be very cautious with these variables (especially the age distribution).
14
Data and Methodology• County level data dating back to 1980/90 to
assess the sources of population growth over two periods, 1990-2000 and 2000-2007 (pre-recession)
• Four samples, Non-metro, Rural, Small Metro (<250,000) and Large Metro (>250,000)
• Extensive sensitivity analysis and staged estimations to mitigate estimation problems associated with multicollinearity or the housing bubble.
15
Non-metro Small MA Large MA1990-2000
2000-2007
1990-2000
2000-2007
1990-2000
2000-2007
Industry Mix Empl. Growth Amenity2 dummy Amenity3 dummy Amenity4 dummy Amenity5 dummy Amenity6 dummy Amenity7 dummy County pop 1990/2000 Pop of nearest or actual UC 1990/2000
4.502***(3.26)0.176(0.90)0.268(1.33)
0.489**(2.34)
0.915***(3.89)
1.146***(3.63)
1.499***(4.40)
-1.7E-06(-1.10)4.1E-07(1.15)
-2.161**(-2.19)0.266(1.18)
0.446**(2.00)
0.520**(2.29)
0.764***(3.12)
0.835***(2.79)0.293(0.90)
6.0E-06***(4.31)
4.3E-07***(2.95)
6.881**(2.11)
0.745***(3.50)
0.789***(3.29)
0.715***(2.62)
1.062**(2.42)0.915*(1.91)0.644(1.02)
3.0E-10(0.00)
1.3E-06(1.34)
0.377(0.12)
0.570***(2.95)
0.487**(2.03)0.210(0.75)0.135(0.32)0.034(0.05)
-1.547**(-2.11)
-1.1E-07(-0.12)
2.7E-06***(3.28)
8.050***(2.69)
0.778**(2.56)
0.634**(2.32)
0.567**(2.15)0.417*(1.81)
0.561***(4.72)
(dropped)
-8.5E-08(-1.46)
4.2E-08*(1.65)
1.523(0.47)
1.528***(3.35)
1.235***(2.73)
1.030**(2.48)
1.223***(2.99)0.589*(1.67)
(dropped)
-1.1E-07(-1.13)2.5E-08(0.81)
Table 1: Dep. Variable: Population growth (%/year) U.S. counties
Notes: Robust t-statistics from STATA cluster command ***, **, and * indicate significance at 1%, 5% and 10%.
16
Non-metro Small MA Large MA1990-2000
2000-2007
1990-2000
2000-2007
1990-2000
2000-2007
Dist. nearest UC
Incr. dist. to a MA
Incr. dist. MA> 250K
Incr. dist. to MA> 500K
Incr. dist. to MA>1.5m
AmenitiesState fixed effectsDemog. vars 1990/2000
-0.010***(-8.22)
-0.004***(-5.60)
-0.003***(-5.60)
-0.002***(-2.85)
-0.001**(-2.56)
YYY
-0.008***(-8.40)
-0.003***(-5.73)
-0.002***(-4.62)
-0.002***(-3.83)
-0.001**(-2.31)
YYY
-0.0002(-0.08)
n.a.
-0.004***(-5.70)
-0.002**(-2.04)
-0.002*(-1.83)
YYY
-0.001(-0.22)
n.a.
-0.003***(-3.02)
-0.003***(-2.62)-0.002*(-1.95)
YYY
0.007**(2.36)n.a.
n.a.
-0.002(-1.54)
0.00008(0.16)
YYY
0.001(0.24)n.a.
n.a.
-0.004**(-2.41)-0.002*(-1.66)
YYY
NR-squared
1,9700.522
1,9700.523
4160.604
4160.516
6410.642
6410.483
F-stats: All dist=0Inc dist to MA=0Amenity vars= 0
17.92***12.55***8.17***
18.04***11.48***8.76***
8.63***11.43***2.39**
3.27***4.36***5.67***
4.79***1.78
5.46***
4.33***5.61***3.49***
Table 1: Dep. Variable: Population growth (%/year) U.S. counties
Notes: Robust t-statistics from STATA cluster command ***, **, and * indicate significance at 1%, 5% and 10%.
17
Interpretation
• Industry Mix Employment growth effect (local growth potential based on industry composition, relative to sources of national growth)—a priori expectation positive, i.e., demand driven migration evident in the 1990-00 period.
• This effect is not statistically significant in the latter period for all FOUR subsamples!
• This appears to be a structural change in economic migration.
Table 2: Pop. Growth in U.S. counties: Impact at Mean ValuesNon-Metro Small MA Large MA
1990-2000
2000-2007
1990-2000
2000-2007
1990-2000
2000-2007
Average pop growth (%/year) 0.595 -0.092 1.266 0.747 1.544 1.094 Ind mix emp gr 1990-00/2000-07 Distance to nearest or actual UC Incremental dist to a metro Inc. dist to metro > 250,000 pop Inc. dist to metro > 500,000 pop Inc. dist to metro > 1.5 mil. pop County pop 1990/2000 Pop of nearest or actual MA1990/2000 County area (sq miles) Amenity2 dummy Amenity3 dummy Amenity4 dummy Amenity5 dummy Amenity6 dummy Amenity7 dummy Great lakes Pacific ocean Atlantic ocean
0.934-0.410-0.225-0.218-0.073-0.079-0.0380.0270.0430.0280.1090.1510.0760.0350.013-0.001-0.007-0.001
-0.127-0.326-0.182-0.133-0.074-0.0640.1470.0310.0380.0420.1820.1610.0640.0250.003-0.007-0.0010.007
1.445-0.004
n.a.-0.389-0.074-0.1200.0000.1800.0380.1130.2830.2530.0690.0400.008-0.007-0.0070.023
0.021-0.016
n.a.-0.235-0.108-0.141-0.0090.4200.0230.0860.1740.0740.0090.001-0.019-0.006-0.007-0.032
1.7300.198n.a.n.a.
-0.0580.009-0.0230.071-0.0140.0700.2680.1890.0330.030n.a.
-0.009-0.005-0.126
0.0850.029n.a.n.a.
-0.127-0.177-0.0330.048-0.0490.1380.5220.3440.0950.031n.a.
-0.017-0.022-0.134
N 1970 1970 416 416 641 641Notes: The table reports the regression coefficient reported in Table 1 multiplied by the variable mean.
19
Further Investigation: LF Part
Dep. Var.: (Empl/Pop18+)00/07 - (Empl/Pop18+)90/00
• Does positive (negative) local demand shocks (decline) generate a higher (lower) employment rate?– The industry mix variable and emp/pop are measured in the same
units.– Industry Mix Growth variable is positive significant in both
periods for all samples, i.e., we observe the expected supply response; But MUCH larger in latter period
• Since this supply response did not come from in-migration, we infer it must have been in the form of increased local (internal) labor force participation response
20
Non-metro Small MA Large MA1990-2000
2000-2007
1990-2000
2000-2007
1990-2000
2000-2007
Industry Mix Empl. Growth
0.200***(4.41)
0.725***(9.65)
0.222***(3.03)
0.547***(5.10)
0.269***(2.95)
0.513***(4.00)
Table 3: Dep. Var.: Diff. in employment-pop (18+year) ratio U.S. counties
Notes: Robust t-statistics from STATA cluster command ***, **, and * indicate significance at 1%, 5% and 10%.
21
What about slack labor markets post-2000?• Perhaps in slack labor markets, more of the labor
demand shock could be satisfied by local labor supply. • The 1990s had strong labor markets, but weaker post
2000. Does this explain our employment/pop results—i.e., it is a cyclical phenomenon when measured at the mean.
• We estimate a series of pop growth quartile regressions at 10, 25, 75, 90th percentiles. In both decades, the slack labor market hypothesis suggests industry mix would have a stronger impact in areas that are growing faster. – We show 75/25 difference in quantile regression results.
22
Non-metro Small MA Large MA1990-2000 2000-2007 1990-2000 2000-2007 1990-2000 2000-2007
Ind mix emp growth rt. 1990-00/2000-07
.522(0.38)
1.22(1.05)
5.73(1.29)
0.92(0.15)
5.78(1.37)
7.70*(1.69)
N 1972 1972 416 416 641 641 .75 Pseudo R2 0.3471 0.3441 0.4506 0.4118 0.4663 0.4067 .25 Pseudo R2 0.3215 0.3560 0.3917 0.3267 0.4132 0.2719
Notes: The coefficients are the difference between the industry mix regression coefficients in a 75th percentile quantile regression model and the corresponding coefficient in the 25th percentile regression model. In parentheses are the bootstrapped t-statistics for the statistical significance between the two estimates using 250 repetitions.
Table 4: Diff. in 75-25 interquantile regression results
23
Other hypotheses• Demographics and super reduced form.
– Due to possible self-sorting we remove the demog variables.– Not shown, the results are basically the same.
• Home ownership and housing bubble. In the bubble, home ownership rates rose. Did this impede migration? – Partridge and Rickman (1997) and Oswald (1997) argue that home
ownership reduces out-migration due to selling ‘costs’. We also argue that it reduces in-migration due to a lower housing supply.
• When adding home ownership share to the model, it supports these migration expectations, though our key results are unchanged.
24
Other hypotheses• Armed forces. Pingle (2007) argues that a decline
in the size of the military and resulting decline in transfers reduced gross migration. Did this affect net migration?– We add the armed forces employment share variable
to the model.– While it has the expected negative coefficient, the
industry mix results are basically unchanged.
25
Table 5: Selected sensitivity analysis regression results
Panel ANon-metro Small MA Large MA
1990-2000 2000-2007 1990-2000 2000-2007 1990-2000 2000-2007
Ind mix emp growth rt. 1990-00/2000-07
4.89***(3.58)
-2.08**(-2.11)
8.06***(2.60)
2.27(0.72)
8.89***(3.13)
3.90(1.24)
Armed Forces Emp Share 1990/2000
-0.045***(-2.72)
-0.024(-1.51)
-0.072***(-3.84)
-0.073***(-2.93)
-0.086***(-5.95)
-0.121***(-5.01)
N 1970 1970 416 416 641 641R2 0.526 0.524 0.630 0.536 0.668 0.520
26
Other hypotheses• Initial housing and labor market disequilibrium.
– Housing bubble.• Regress for the initial period 1990 and 2000 log median housing
price and median wage on quality and other site-specific characteristics (Clark et al., 2003; Partridge et al., 2010).
• Ln(median housing price) = βhXh + eh
• Ln(median wage) = βwXw + ew
• Use the residuals eh & ew as measures of initial ‘disequilibrium’ and place them in the base regression model. This allows us to assess if a frothy housing market affects our results.– We find that our results are unchanged for industry mix and that a initial
local housing bubble is positively associated with net migration. (also controlling for 1980 wages and housing prices)
27
Panel B
Non-metro Small MA Large MA1990-2000 2000-2007 1990-2000 2000-2007 1990-2000 2000-2007
Ind mix emp growth rt. 1990-00/2000-07
5.28***(3.47)
-2.48**(-2.24)
7.72**(2.18)
0.61(0.18)
9.28***(2.88)
1.79(0.46)
Resids from log(avg. wage) 1990/2000
-0.078(-0.25)
0.40(1.19)
-0.17(-0.18)
1.22(1.08)
0.55(0.94)
1.88**(2.29)
Resids from log(avg. rent) 1990/2000
1.40***(4.55)
1.61***(5.49)
1.51**(2.55)
1.16(1.22)
1.08**(2.38)
0.32(0.36)
N 1970 1970 416 416 641 641R2 0.5326 0.5367 0.6107 0.5215 0.6470 0.4899
Table 5: Selected sensitivity analysis regression results
28
Other hypotheses
• Demographics– An aging population may alter migration patterns
• Yet, this is in the constant term and in the state fixed effects. Moreover, in terms of life cycle effects, it would alter how the amenity variables affect migration.
• Directly controlling for the age composition has endogeneity issues—i.e., the young move to growing places.
• When we control for age composition, the results are fairly similar. We take the results cautiously because of the self-sorting makes age composition endogenous.
29
Panel CNon-metro Small MA Large MA
1990-2000 2000-2007 1990-2000 2000-2007 1990-2000 2000-2007Ind mix emp growth rt. 1990-00/2000-07
6.86***(5.38)
-0.77(-0.66)
8.37***(3.08)
2.35(0.77)
11.44***(4.40)
8.16**(2.34)
Pop share 7-17 -0.163***(-4.62)
-0.16***(-4.78)
-0.206*(-1.85)
-0.30***(-2.69)
-0.103(-0.96)
-0.50***(-4.00)
Pop share 18-24 -0.070**(-2.28)
-0.09***(-3.05)
-0.092(-1.14)
-0.117(-1.34)
0.017(0.25)
-0.296***(-3.17)
Pop share 25-54 -0.034(-1.12)
-0.12***(-3.95)
-0.102(-1.18)
-0.1448(-1.46)
0.093(1.29)
-0.240**(-2.06)
Pop share 55-59 -0.060(-1.19)
-0.08*(-1.82)
-0.292**(-2.04)
-0.308**(-2.02)
-0.41***(-3.57)
-0.583***(-5.70)
Pop share 60-64 0.071(1.40)
-0.054(-1.19)
-0.041(0.29)
-0.2051*(-1.65)
-0.177(-1.12)
-0.485***(-3.80)
Pop share 65+-0.141***
(-4.82)
-0.184***
(-6.86)-0.190**(-2.13)
-0.24**(-2.60)
-0.028(-0.40)
-0.354***(-3.62)
N 1970 1970 416 416 641 641R2 0.5636 0.5739 0.6402 0.5876 0.7014 0.5625
Table 5: Selected sensitivity analysis regression results
30
Further Investigation• Add occupation mix variable to assess whether mobility became
more linked with skill/occupation shocks, implying increased mobility across industries while staying in the same occupation.– Statistically significant results suggest weak evidence of this being true
post-2000, but not during the 1990s.
• Net-Migration—much like Pop. Chg.– Industry Mix variable strongly positive in initial period, insignificant in
the latter period all samples– Modest reduction in Amenity influence between periods for Small and
Large Metros
• Net migration models with international migration are also not qualitatively different
• Using in- and out-migration flows do not alter our conclusions.
31
Summary and Conclusions• Preliminary results suggest that one culprit of the decline in
migration may be a modest decline in amenity migration. Rural/urban migration trends continue as well.
• The major factor, however, appears to be a decline in economic migration due to differential local economic growth
• Until this decade, a major factor underlying migration was job related, but local labor force demand appears to be increasingly satisfied with local and nearby workers rather than new migrants
• We conclude that diminishing migration is not reflective of the U.S. reaching a spatial equilibrium. Rather there is a structural change in economic migration.
32
Conclusions• U.S. labor markets took on more of a European flavor with local
labor demand being satisfied with local labor supply.• Does this imply that the U.S. labor market is less flexible due to less
geographic mobility?• Perhaps not….• Technological change facilitates more information about other labor markets
but it also helps one find out about information about local job opportunities (much like unexpected NEG and transport results or internet is a complement to face-to-face interaction).
• There is evidence of more industry mobility (Kambourov and Manovskii, 2008).
• There is also more temporary and contingent workers.• Together, it may be less necessary to relocate for economic reasons.
Geographical mobility is being replaced with industry mobility. This needs to be addressed with micro data.
33
Thank-you
34
Appendix slides
Annual Gross Migration Rates, U.S. 1947-2008
-
2.0
4.0
6.0
8.0
10.0
12.0
14.0
16.0
18.0
20.0
22.0
Percent
Total moved Moved to diff county Moved to diff county diff state
Source: U.S. Census Bureau, Current Population Survey.
36
1990-2000 period 2000-2007 periodNon-metro Rural Sm. MA Lg. MA Non-metro Rural Sm. MA Lg. MA
Pop growth 1990-00 and 2000-07
0.5950(1.08)
0.4776(1.10)
1.2661(1.12)
1.5442(1.36)
-0.0919(0.96)
-0.2795(0.95)
0.7472(1.07)
1.0945(1.52)
Chg in emp/pop. (18+) ratio 1990-00/2000-07
0.0165(0.05)
0.0173(0.05)
0.0191(0.03)
0.0153(0.03)
-0.0027(0.05)
-0.0003(0.06)
-0.0121(0.03)
-0.0257(0.03)
Ind mix emp gr 1990-00/2000-07
0.2075(0.04)
0.2095(0.04)
0.2100(0.03)
0.2149(0.02)
0.0588(0.03)
0.0617(0.03)
0.0564(0.03)
0.0561(0.02)
Distance to nearest or actual UC
41.07(36.52)
59.91(30.56)
16.85(17.00)
26.27(16.77)
41.07(36.52)
59.91(30.56)
16.85(17.00)
26.27(16.77)
Incremental dist to a metro 55.40(51.67)
43.47(49.93)
0.00(0.00)
2.33(7.37)
55.40(51.67)
43.47(49.93)
0.00(0.00)
2.33(7.37)
Inc. dist to metro > 250,000 pop
66.80(106.20)
76.02(115.19)
93.23(93.26)
0.00(0.00)
66.80(106.20)
76.02(115.19)
93.23(93.26)
0.00(0.00)
Inc. dist to metro > 500,000 pop
42.89(66.07)
45.32(68.95)
36.89(59.07)
36.29(73.34)
42.89(66.07)
45.32(68.95)
36.89(59.07)
36.29(73.34)
Inc. dist to metro > 1,500,000 pop
89.03(111.10)
83.45(106.24)
78.54(115.44)
99.37(139.88)
89.03(111.10)
83.45(106.24)
78.54(115.44)
99.37(139.88)
County pop 1990/2000 22442(20585)
13770(10401)
72161(64892)
270700(539956)
24441(22808)
14832(11427)
82750(76200)
308194(595249)
Pop of nearest or actual MA 1990/2000
65459(93944)
73970(113164)
133332(49192)
1486906(2726624)
72664(110160)
82460(132956)
151186(59500)
1681592(2997111)
% 1990(00) Pop. African American
7.7560(14.74)
7.1534(14.77)
8.9211(12.25)
11.0011(13.98)
7.8995(14.89)
7.3009(14.87)
8.8893(12.24)
11.2831(14.18)
% 1990(00) Pop. Native America
1.8192(6.72)
2.0982(7.66)
0.8659(2.67)
0.6514(1.64)
1.9445(7.06)
2.2666(8.09)
0.8963(2.63)
0.6623(1.49)
% 1990(00) Pop. Hispanic 4.3354(11.64)
4.2174(11.46)
3.7863(9.17)
4.8435(9.82)
5.9260(12.55)
5.5656(12.21)
5.7645(10.66)
7.0098(11.25)
% 1990(00) Pop. Asian 0.3159(0.43)
0.2164(0.27)
0.7911(1.16)
1.3257(2.30)
0.4241(0.46)
0.3173(0.31)
1.0180(1.28)
1.8556(2.93)
% 1990(00) Pop. Other origin 1.7779(4.84)
1.7162(4.85)
1.6347(3.95)
1.9858(4.06)
2.4443(4.88)
2.2584(4.71)
2.5950(4.98)
2.9430(4.72)
% 1990(00) High School Grad. 35.00(5.96)
35.25(5.82)
34.26(6.32)
32.47(6.07)
35.97(5.89)
36.40(5.59)
34.27(6.76)
31.49(7.00)
% 1990(00) Pop. 25+ Some College
15.65(4.38)
15.28(4.32)
17.10(4.41)
18.19(4.34)
20.04(4.52)
19.98(4.62)
20.86(3.99)
21.22(3.82)
% 1990(00) Pop. 25+ Assoc. Degree
5.15(2.20)
5.01(2.26)
5.56(1.96)
5.79(1.77)
5.47(2.05)
5.29(2.02)
5.97(1.86)
6.23(1.68)
% 1990(00) Pop. 25+ College Degr
11.75(4.73)
10.98(4.12)
14.83(6.98)
17.68(8.27)
14.32(5.64)
13.51(4.99)
18.01(8.04)
21.90(9.64)
% 1990(00) Pop. that is female 51.02(1.63)
50.95(1.65)
50.97(1.52)
51.05(1.60)
50.37(2.07)
50.25(2.18)
50.63(1.55)
50.80(1.45)
% 1990(00) Pop. married 59.92(5.91)
60.79(5.77)
58.03(5.91)
57.36(6.89)
58.14(5.15)
58.83(5.05)
56.95(5.29)
56.83(6.09)
% 1990(00) Pop. with a disability
10.05(3.05)
10.29(3.28)
9.11(2.30)
8.28(2.20)
12.25(3.33)
12.39(3.43)
11.50(3.01)
11.33(2.75)
N 1972 1300 416 641 1972 1300 416 641
Appendix Table 1: Descriptive Statistics, Mean and (Standard Deviation) for U.S. Counties, Selected Variables
37
Impact at one std. dev. Pop. Change ModelNon-metro Rural
1990-2000 2000-2007 1990-2000 2000-2007Average pop growth (%/year) 0.595 -0.092 0.478 -0.279Ind mix emp gr 1990-00/2000-07Distance to nearest or actual UCIncremental dist to a metroInc. dist to metro > 250,000 popInc. dist to metro > 500,000 popInc. dist to metro > 1,500,000 popCounty pop 1990/2000Pop of nearest or actual UC 1990/2000County area (sq miles)Amenity2 dummyAmenity3 dummyAmenity4 dummyAmenity5 dummyAmenity6 dummyAmenity7 dummyGreat lakesPacific oceanAtlantic ocean
0.159-0.365-0.210-0.346-0.112-0.099-0.0350.0380.0540.0640.1320.2260.2530.1970.139-0.009-0.074-0.004
-0.069-0.290-0.170-0.211-0.114-0.0790.1370.0470.0470.0970.2190.2400.2110.1430.027-0.040-0.0080.036
0.869-0.523-0.178-0.247-0.084-0.0960.1030.012-0.0260.0000.0170.0860.0790.0260.0090.000-0.004-0.002
-0.148-0.410-0.131-0.174-0.081-0.0760.3060.020-0.005-0.0060.0570.0620.0510.0140.002-0.002-0.0010.003
N 1970 1970 1300 1300
38
Impact at one std. dev. Pop. Change ModelSmall MA Large MA
1990-2000 2000-2007 1990-2000 2000-2007Average pop growth (%/year) 1.266 0.747 1.544 1.094Ind mix emp gr 1990-00/2000-07Distance to nearest or actual UCIncremental dist to a metroInc. dist to metro > 250,000 popInc. dist to metro > 500,000 popInc. dist to metro > 1,500,000 popCounty pop 1990/2000Pop of nearest or actual UC 1990/2000County area (sq miles)Amenity2 dummyAmenity3 dummyAmenity4 dummyAmenity5 dummyAmenity6 dummyAmenity7 dummyGreat lakesPacific oceanAtlantic ocean
0.190-0.004
n.a.-0.389-0.118-0.1760.0000.0590.0520.2670.3790.3420.2620.1860.070-0.033-0.0480.070
0.010-0.016
n.a.-0.235-0.173-0.208-0.0080.1470.0320.2040.2340.1000.0330.007-0.169-0.026-0.044-0.096
1.7300.198n.a.n.a.
-0.0580.009-0.0230.071-0.0140.0700.2680.1890.0330.030n.a.
-0.009-0.005-0.126
0.0850.029n.a.n.a.
-0.127-0.177-0.0330.048-0.0490.1380.5220.3440.0950.031n.a.
-0.017-0.022-0.134
N 416 416 641 641
39
Background cont’d.“Slump Creates Lack of Mobility for Americans,”
New York Times by Sam Roberts, April 23, 2009 “the number of people who changed residences declined
to 35.2 million from March 2007 to March 2008, the lowest number since 1962, when the nation had 120 million fewer people.”
Refers to concern that “if job-related moves are getting suppressed and workers are not getting re-sorted to the jobs that best use their skills,” long term negative consequences.
Reasons for greater immobility: greater home ownership, two earner households, ageing (LF).
40
Theoretical Model cont’d.• Households in region i derive utility from
consumption (X), housing (H), and site-specific amenities (S).
• HH provide labour to the market for which they receive wages (w) with some probability of employment (e), and pay rents (r) for housing
• HHs in region i are located in a spatial environment such that they can access employment opportunities, built amenities and consumption goods in n other regions j by traversing Distance (Dij)
41
Theoretical Model cont’d.• HH and Firm expected conditions location i
– Indirect Utility Function: Vi(wi, ri
H, ei, Si H, DISTij, ·)
– Indirect profit function:Πi(wi, rF
i, mi, Si F DISTij, ·)
• LR equilibrium requires HH and Firm re-locations, at rate α, to equalize expected utility and profits– HHNMi=αH
i(Vj-Vi) for any j; HH net migration– FNMi= αF
i(COSTj-COSTi) for any j; firm net migration
• Pop. Chg.i = f(initial conditions in i incl. DISTij)
42
Descriptive Statistics Average Annual
Population Growth (%)Average Chg in Empl/Pop
18+ ratio
1990-2000 2000-2007 1990-2000 2000-2007
Non-Metro 0.595 -0.092 0.0165 -0.0027
Rural 0.478 -0.279 0.0173 -0.0003
Small Metro 1.266 0.747 0.0191 -0.0121
Large Metro 1.544 1.094 0.0153 -0.0257
Phoenix
Salt Lake City
Provo-Orem
Cedar City
St. George
88 km
88 km58 km
146 km
132 km
278 km
43 km
321 km
156 km
ARIZONA
UTAH
Garfield county,Utah
Garfield County, a rural Utah county, ≈ 4,000 residents (1990). The nearest urban area is Cedar City (a MICRO) located 88kms away. The nearest MA is St. George (≈ 90,000 pop.), 146kms away, an incremental distance of 58kms (146-88). Nearest larger MA > 250K, which is Provo-Orem, UT (pop. of 377,000), is 278kms from Garfield County, incremental distance versus St. George is 132kms (278-146). The nearest MA > 500K, the next higher tier, Salt Lake City, UT (969,000 people). Salt Lake is 321kms from Garfield County, incremental distance of 43kms (321-278). Nearest MA > 1.5 million people, the next higher tier above Salt Lake, is Phoenix, AZ (3.25 million 1990 pop.). Phoenix is 477kms away from Garfield County, an incremental distance of 156kms (477-321).
Distance calculations
43
44
Casper
Laramie
Fort Collins
Denver-Aurora
Carboncounty
129km
129km
14km
143km
67km210km
72km
WYOMING
COLORADO
Representing the Urban HierarchyFor Carbon County, a rural Wyoming county (pop. 8,500) nearest urban area is Albany County (a MICRO), containing the city of Laramie, 129kms away. The nearest MA is Casper (67,000pop.), located 143kms away, an incremental distance of 14kms (143-129). Because Casper is a small MA, Carbon County will also be influenced by its remoteness from larger MA of at least 250K. Nearest of >250 is Ft. Collins, Colorado, (pop. 250,000) 210kms from Carbon County, but the incremental distance beyond Casper is 67kms (210-143). The nearest MA of at least 500K, the next tier, is Denver, Colorado (2.5 pop). Denver is 282kms away from Carbon County, an incremental distance of 72kms (282-210). Because Denver is already over 1.5 million population, no incremental distance to reach a MA of at least 1.5 million.