1
Why is U.S. Poverty Higher in Nonmetropolitan
than in Metropolitan Areas?

• by Monica Fisher, OSU AREC
• In Growth and Change,
• (March 207)
• Vol. 38 No. 1 pp. 56-76
• Presented by DAnne Hammond

2
Rural/Urban Poverty, Fisher

• Examines whether the difference in poverty levels
  reflects personal choice in addition to
  structural explanations of limited economic and
  social opportunities.
• i.e. structural condition and sorting hypothesis?
  

3
The question

• ...asking if the disproportionate poverty in
  non-metro places partly reflects attitudes of
  people with personal attributes related to
  poverty they may be attracted to non-metro
  places or otherwise reluctant (or unable) to
  leave them.

4
Observable differences in poverty rates

• Other research shows the odds of being poor are
  1.2 to 2.3 times higher in rural areas
• 1/20 metro counties poverty rate 20 or higher
• 1/5 non-metro counties poverty rate 20 or higher

5
Theoretical model

• y f (age female/not white/not education
  unemployed/not in labor force/not retired
  disabled married household size young child
  present/not non-metro/not)
• where y adj. income-needs ratio

6
Method

• Series of multivariate regression models using
  OLS to model poverty across place
• All models are variations of
• y1 a0 a1xi a2ni a3si ei
• Where y pretax income-to-need need is
  census-based poverty threshold
• x individual factors race, age, gender,
  presence of children, etc.
• n binary variable indicating non-metro
• s fixed-effects controlling for state-level
  expenditures, tax structure, etc.
• e error term, assumed to be i.i.d.

7
Variable y

• Dependent variable y (income-to-need) is adjusted
  for housing cost differences using fair market
  rent values (FMR)
• Persistent differences in housing costs between
  metro and non-metro

8
Data source

• Panel Study of Income Dynamics (PSID)
• Nationally representative sample
• Longitudinal data
• Survey following apx. 5,000 families since 1968
• This research uses the nine panels between 1985
  and 1993 that include non-metro variable

9
Data subsample

• Household min. 2 yrs. observations in study
• Restricted to lower income distribution
• Records must have complete data for all variables
• Head of household is a proxy for entire household
• Result is sample of 2,007 household heads in
  poverty in 1993 and at least one other year
  between 1985 and 1993
• Average number of years in sample 7

10
Final data set characteristics

• Differences (statistically significant) between
  whole sample and selected sub-sample
• Female
• With young child
• Non-metro area
• More likely non-white and unemployed
• Lower education levels

11
Results

• F-stat indicates joint significance of
  explanatory variables (95 ci)
• Indirect evidence supporting both structural
  condition hypothesis and self-sorting to
  non-metro
• Metro to non-metro movers have increased
  income-needs ratio of 25
• Non-metro households economically worse off
  (ceteris parabus)

12
Policy Implications

• Further empirical studies on place-level and
  individual-level variables over time
• ...can improve the design of anti-poverty
  policy, providing insights on what combinations
  of human-capital and community-strengthening
  policies are most likely to reduce non-metro
  poverty and its unfavorable consequences. (p.
  73)