Harvard Government 90dn Lecture 8

Government 90dn Mapping the Census Lecture 8: Spatial Analysis; Economic Data in the Census

Sumeeta Srinivasan [email protected]

Outline ƒ Intro to Spatial Analysis ƒ MAUP ƒ Using Economic data from the Census

Spatial Analysis ƒ ƒ ƒ ƒ ƒ ƒ

Query Measurements Transformation Descriptive Summary Optimization Hypothesis Testing

Attribute Queries with Crashes

Attribute Queries with Crashes

Location Queries with Crashes

Location Queries with Crashes (and fatalities)

Queries with Crashes (and fatalities)

Crash Statistics ƒ Near high speed roads ƒ Total non-fatality incidents 60 (1 fatal) ƒ Average non-fatality incidents: 0.47

ƒ On all roads ƒ Total non-fatality incidents 711 (4 fatal) ƒ Average non-fatality incidents: 0.27

ƒ On low speed roads ƒ Total non-fatality incidents 651 (3 fatal) ƒ Average non-fatality incidents: 0.26

Measurement ƒ Distance and length ƒ Shape

1991 and 2001 Congressional District Boundaries (San Diego, CA)

Transformations ƒ Buffers ƒ Overlay

Buffers (Discrete vs Continuous)

Dissolve CC CC CC CC

RR RR CC CC

CC CC CC CC

#

Item

1

C

2

R

3

C

4

I

II II CC RR

R

I

C R

Dissolve Example

Intersect ƒ Use Intersect when you want to overlay a layer with the polygons in another layer to get only the overlap

Intersect ƒ Flood zones that intersect parking

Intersect Result

Descriptive Summaries: Centroids

Hypotheses Testing

Moran Coefficient (from Rick Glazier and Peter Gozdyra, University of Toronto ) Statistic\Value

-1.0

0.0

1.0

Moran Coefficient

Strong negative autocorrelation

Random distribution of values

Strong positive autocorrelation

Strong positive autocorrelation

Random distribution of values

Geary Ratio

2.0

Strong negative autocorrelation

Moran?

Hypotheses testing with Crashes

Moran of Crashes in Cambridge Moran's Index = -0.131211 Expected Index = -0.001319 Variance = 0.115055 Z Score = -0.382937

Local Moran of Crashes

MAUP (Modifiable Areal Unit Problem) ƒ Ecological Fallacy, Robinson, 1950 ƒ 1930 census county data, r=.773 ƒ 1930 census individual data, r=.203 (correlation between being black and being illiterate) ƒ Iowa Study, Openshaw and Taylor, 1979 ƒ 99 counties in Iowa, r=.35 ƒ Regroup into 48 regions many times, -.55 < r < .89 ƒ Regroup into 12 regions many times, -.94 < r < .99 (correlation between % 65+ and % registered Republican)

MAUP example

Economic Census ƒ Conducted every five years, in years ending in ‘2’ and ‘7’ ƒ Data from the 2002 Economic Census are currently being released

Economic Census ƒ Industry Series ƒ Not useful to map

ƒ Subject series ƒ Specialized

ƒ Geography series ƒ Place and ZIP code levels

The Base Multiplier ƒ Base Multiplier can be expressed as a ratio: BM = Total Employment Basic Employment ƒ Using this approach, analysts can project impacts upon the total economy from expected changes to the basic sector ƒ Assumed that the ratio of total local employment activity to basic employment (the BM) does not vary over time.

Leon County’s Base Multiplier

Employment by Sector in Leon County, 1999 Source: Florida Department of Labor and Employment Security Sector Units Employees Percentage Se rvice s

3,177

37,433

26.8%

Re ta il Tra de

1,320

24,147

17.3%

FIRE

664

6,260

4.5%

Co nstruction

700

5,695

4.1%

Tra ns, Comm And P ublic Util

214

3,818

2.7%

W hole sa le Tra de

427

3,764

2.7%

Ma nufa cturing

172

3,022

2.2%

Agriculture , Fore stry & Fishing

152

1,038

0.7%

Sta te Gove rnme nt

208

41,122

29.5%

Loca l Go ve rnme nt

24

11,070

7.9%

36 7,303

1,677 139,492

1.2% 100%

Fe de ra l Go ve rnme nt LEO N CO UNTY TO TAL

Basic Sec tor Employment Non- Basic Sec tor Employment T otal Employment Base Multiplier

46,859 92,633 139,492 2.98

(=139,392/46,859)

Four steps in EB Analysis 1. Area to be Studied (Geography) 2. Unit of Analysis (Measure of the Economy) 3. Data to be Used (Source for Input Data) 4. Technique to be Used (Analytical Methods)

1. Choosing a Study Area ƒ County: The most commonly used study area because of excellent data availability ƒ MSA: Best unit for urban analysis; Built on counties, so excellent data availability as well ƒ Economic Region: A shopping area or media area is useful, but poor data availability makes this a rarely used analysis area ƒ State: A study area that is too aggregated and likely to undercount basic sector activity

2. Selecting the Unit of Analysis ƒ ƒ ƒ ƒ

Employment: The number of jobs by industry Payroll: Annual payroll for firms by industry Sales: Dollar sales by industry Value Added: Like sales, but eliminates double-counting by subtracting a firm’s purchases from their sales

3. Selecting the Data Set County Business Patterns: ƒ Pros: Available annually; includes employment, payroll, sales ƒ Cons: Derived from a combination of sources; Does not include Government employment Economic Census ƒ Pros: Contains employment, payroll, sales ƒ Cons: Collected only every five years but not available until several years later ES202 Data ƒ Pros: Available annually and by quarter; Includes employment and payroll ƒ Cons: Not always available for all areas

4. Economic Base Analysis Techniques ƒ Direct Method: The simplest and most straightforward, this approach assumes that certain industries are Basic or Non-basic ƒ Location Quotients: Related to the concentration concept, this technique determines the local share of an industry

The Direct Approach ƒ Assigns activities to the Basic and Non-basic sectors on the basis of assumed sales patterns for different types of industries. ƒ Sectors typically assigned to the Basic sector: Manufacturing, State/Federal Government, Agriculture, Forestry, Fishing, Hotels/Lodging, Mining. ƒ Sectors typically assigned to the Non-Basic sector: Retail Trade, Local Government, Wholesale Trade, Services, Transportation, Commercial, Utility, Construction.

The Location Quotient Technique ƒ Location quotients compare the local share of a given industry to the share of that industry for a larger area ƒ The formula: LQi = eit/eTt Eit/ETt where: eit = Local employment in sector i at time t eTt = Total local employment at time t Eit = National employment in sector i at time t Ett = Total national employment at time t ƒ Three values are possible: 1) Industries with LQ’s = 1 (Self-Sufficiency) 2) Industries with LQ’s < 1 (Net Importer) 3) Industries with LQ’s > 1 (Net Exporter)

Calculating a Location Quotient for Massachusetts Medical related employment MA's employment in Medical related MA's total employment US employment in Medical related US total employment MA share of employment

404,200 3,323,200 6,779,990 137,632,000 404,200

in Medical related

3,323,200

US share of employment

6,779,990

in Medical related

137,632,000

MA concentration US concentration

12% 5%

Location quotient

2.40

12%

5%

Calculating a Location Quotient for Massachusetts Manufacturing Industry MA's employment in Manufacturing MA's total employment US employment in Manufacturing US total employment MA share of employment in Manufacturing US share of employment in Manufacturing

407900 3,323,200 17263000 137,632,000 407900

12%

3,323,200 17,263,000 137,632,000

MA concentration US concentration

12% 13%

Location quotient

0.98

13%

BLS Location Quotients

County Business Patterns

Query to get only County data for Massachusetts

Query to get only data for Massachusetts where state and county data are available

Query to get only Middlesex County (17) data for Massachusetts (25) for codes 621-623 Health care services

Group by counties

Group by counties

Group by States

Location Quotients ƒ Middlesex and Suffolk county’s share in health NAIC 621-623 related employment versus state share in health related or state overall employment

Location Quotients by County in MA (compared to employment in health only)