1. Economic Growth IN THE UNITED STATES OF AMERICA A County-level Analysis
April Harris
Elana Kaufman
Sohair Omar
Elizabeth Pearson

2. Objective
To explore the factors driving differences in regional economic growth across the United States.
To replicate the analysis in the OECD paper, “The Sources of Economic Growth in OECD Regions: A Parametric Analysis,” (December 2008) for the U.S. case.

3. Agenda Theory Data Summary Statistics Results Findings/Conclusion
Future research/Recommendations
Questions

4. What theories explain economic growth?
Neo-Classical Theory
Endogenous Growth Theory
New Economic Geography (NEG)

5. Neo-Classical Theory Assumes Diminishing Returns And Exogenous Technology
Key assumptions:
Capital is subject to diminishing returns
Perfect competition
An exogenously determined constant rate reflects the progress made in technology
3 Key factors:
Capital intensities
Human capital
Technology (not included in the model; exogenous)
Key assumptions:
Capital is subject to diminishing returns
Perfect competition
An exogenously determined constant rate reflects the progress made in technology
Three key factors:
Capital intensities
Human capital
Technology (not included in the model; exogenous)
Improvements: Endogenous growth considered that returns to capital do not diminish because human capital entails knowledge spillovers and external benefits

6. Neo-Classical Theory Predicts Convergence
Long-run growth is the result of continuous technological progress, which is determined exogenously
Key implication: Conditional convergence
Problems
Limited empirical evidence of convergence
Leaves technological progress out of the model
Key assumptions:
Capital is subject to diminishing returns
Perfect competition
An exogenously determined constant rate reflects the progress made in technology
Three key factors:
Capital intensities
Human capital
Technology (not included in the model; exogenous)
Improvements: Endogenous growth considered that returns to capital do not diminish because human capital entails knowledge spillovers and external benefits

7. Endogenous Growth Theory Assumes Diminishing Returns and Endogenous Technology
Key assumptions:
Capital is subject to diminishing returns
In many endogenous growth models the assumption of perfect competition is relaxed, and some degree of monopoly power is thought to exist.
3 Key factors:
Physical capital
Human capital
Technology (included in the model: endogenous)

8. Endogenous Growth Theory: Internal factors are the main sources of economic growth
Investing in human capital  the development of new forms of technology & efficient and effective means of production  economic growth
Investment in human capital (education and training of the workforce) is an essential ingredient of growth
The main implication: policies which embrace openness, competition, change and innovation will promote growth.
Theory emphasizes that private investment in R&D is the central source of technical progress
No convergence is predicted.
The main implication of recent growth theory is that policies which embrace openness, competition, change and innovation will promote growth. Conversely, policies which have the effect of restricting or slowing change by protecting or favouring particular industries or firms are likely over time to slow growth to the disadvantage of the community

9. New Economic Geography: Why is manufacturing concentrated in a few regions?
Economic geography: the location of factors of production in space
Key factors
Economies of scale
Transportation costs
Location of demand
Population
What does NEG seek to answer?
Why and when does manufacturing become concentrated in a few regions, leaving others relatively undeveloped?
In order to realize scale economies while minimizing transport costs, manufacturing firms tend to locate in the region with larger demand, but the location of demand itself depends on the distribution of manufacturing. Emergence of a core-periphery pattern depends on transportation costs, economies of scale, and the share of manufacturing in national income.
How does NEG theory answers these questions?
Through a model of geographical concentration of manufacturing based on the interaction of economies of scale with transportation costs
Concentration of manufacturing in one location need not always happen and that whether it does depends in an interesting way on a few key parameters
If one region has right mix of key factors before - even just slightly before - another similar region, the leading region takes off and the other may not grow.

10. New Economic Geography predicts that the right mix of key factors causes growth
Key implications
Agglomeration raises wages in the core region relative to the periphery
Despite early similarity regions can become quite different
Large nearby demand causes more growth
Producing near one’s main market minimizes transportation costs

11. How does NEG differ from Neo-Classical and Endogenous Growth Theories?
NEG takes scale into account
NEG models propose that external increasing returns to scale incentivize agglomeration
Agglomeration captures, via scale effects, how small initial differences cause large growth differentials over time

12. We obtained data on 3,079 counties between 1998-2007
Variable
Source
Year(s)
Annualized per capita personal income growth
Bureau of Economic Analysis
Log of income in the initial year
1998
Physical capital/infrastructure
ESRI Data and Maps 9.3 Media Kit
2008
Education rates
U.S. Census
2000
Innovation Index
Economic Development Administration
Employment rate
Employment specialization
Census of Employment and Wages
Accessibility to Markets/Distance to Markets

13. Per Capita Personal Income
Ranges from $8,579 in Loup County, NE to $132,728 in Teton County, WY
Used to create three variables:
Dependent variable: annualized per capita personal income growth 1/10 * ln(income in 2007) – ln(income in 1998)
Highest: 7% in Sublette, WY
Lowest: -3% in Crowley, CO
Mean: 1%
Independent variable: log of income in the initial year, 1998
Highest: $76,450 in New York, NY
Lowest: $7,756 in Loup, NE
Independent variable: per capita personal income in nearby counties, weighted by distance and other spatial measures
Personal income includes net earnings (earnings by place of work less contributions for government social insurance), rental income, personal dividend income, personal interest income, and personal current transfer receipts
Per capita personal income is the personal income of residents of a given area divided by the resident population of the area, using the Census Bureau’s annual midyear population estimates

14. Cost of Living?

16. Growth rates look low. Investigate them. -- PRW

17. Infrastructure A measure of Physical Capital.
Mileage of major roads by county
Airports by county

18. Major Road Mileage by County

19. Number of airports by County

20. Education Rates Source: 2000 Census
Percent of population with less than high school degree
Highest: 62.5% in Starr, TX
Lowest: 4.4% in Douglas, CO
Median: 21.6%
Percent of population with a high school diploma
Highest: 53.5% in Carroll, OH
Lowest: 12.4% in Arlington, VA
Median: 34.7%
Percent of population with more than a high school degree
Highest: 82.1% in Los Alamos, NM
Lowest: 17.2% in McDowell, WV
Median: 41.4%
These three variables add up to 1
(Capture above info in bar graph)
Signaling instead of human capital
Kelly Becker, showing that when there are barriers to entering university, people pool around the high school diploma
- Tyler, Murnane, and Willet (2000) showing that people with GEDs are paid the same no matter how many years it tok them to get it
percentlessthanhs is slightly positively correlated with percenthsdiploma
percentlessthanhs is very negatively correlated with percentmorethanhs
percenthsdiploma is very negatively correlated with percentmorethanhs
but percentlessthanhs has a positive coefficient on it—does that mean that having fewer HS graduates means you have a higher growth rate? That seems implausible.
regressing totalpcpigrowth on percdentlessthanhs and percenthsdiploma has a negative coefficient on both, but it’s even more negative on percenthsdiploma
Denominator: total population. This may have been a bad idea, because this also means we’re counting children? More children could mean more growth.

21. Innovation Index
[COMING SOON]

22. Employment Rate Source: 2000 Census (for cross-section)
Youth employment rate: population aged 16 – 20 that is working divided by total population 16 – 20
Highest: 100% in Loving, TX
Lowest: 8.78% in Shannon, SD
Median: 46.2%
Working age employment rate: population aged 21 – 65 that is working divided by total population 21 – 65
Highest: 88.4% in Stanley, SD
Lowest: 35.9% in McDowell, WV
Median: 73%
Total employment rate
Highest: 86.7% in Stanley, SD
Lowest: 33.6% in McDowell, WV
Median: 69.9%
(NEED BAR GRAPH!)
Youth employment rate generally has a negative coefficient for growth
working age employment rate has a positive coefficient
check the correlation coefficients

23. Employment Specialization: Measure of Industrial Concentration of Region
Meant to capture notion of agglomeration
The spatial concentration of industry
A determinant of economic growth in NEG growth theory.
How is it modeled?
Specialization indices
Herfindahl Index
Krugman Index

24. Herfindahl Index
The Herfindahl index is the sum across industrial sectors of the square of that sector’s share of employment
Ranges from 0 to 1.0
0 = large number of very small firms
(perfect competition)
1 = a single monopolistic producer
(complete monopoly by a single firm)

25. Krugman Index KI = ∑j|aij-b-ij|
a = the share of industry j in county i’s total employment
b = the share of the same industry in the employment of all other counties, -i
KI = the absolute values of the difference between these shares, summed over all industries
Ranges from 0 to 2.0
0 = county i has industrial composition identical to its comparison counties
2 = county i has industrial composition without any similarity (no common industries) to its comparison counties

26. To Measure Employment Specialization We Chose the Krugman Index
Pros
Cons
Herfindahl Index
Captures industrial specialization
Is an absolute measure; Does not take neighbors into account
Krugman Index
Captures industrial specialization.
Is a relative measure; Compares to one’s neighbors
Argued that the absolute size of
clusters should be the basis for
calculation.
Objections hold that this level is
systematically underestimated for
larger metropolitan areas when
relative levels of concentration are
used.

27. Accessibility to Markets/Distance to Markets
[PENDING]

28. OLS Results
Negative relationship. Evidence of convergence.

29. Scatter Plot
Proof of “crappy” R-squared! Delete.

30. OLS Results
More highways, more growth.

31. OLS Results
Percentages sum to one. Choose which two to include. Conduct a f-test on group of variables.

32. OLS Results

33. OLS Results
F-test on infrastructure variables to see if pair is significant. Do a correlation matrix.

34. OLS Results

35. OLS Results
Correlated? Do a correlation matrix. Do an f-test.

36. OLS Results

37. OLS Results

38. OLS Results

39. OLS Results

40. OLS Results

41. OLS Results

42. OLS Results

43. Modeling Spatial Relationships
Inverse Distance …
K-Nearest Neighbor
Contiguity

44. Contiguous Counties

45. The average county has 5 to 6 neighbors (main point)
How many neighbors does the…

46. Global Spatial Autocorrelation
Growth rates display spatial dependence…Moran’s I…Null hypothesis

47. Own growth rates depend on neighbors (idea)

48. Main Findings

49. Future Research

50. Questions