Opportunity Mapping Methodology 2019

Opportunity Methodology

This page describes the (1) data sources, (2) indicators, and (3) methodology employed in generating the 2019 Opportunity Index for Connecticut.

This analysis updates the 2014 Opportunity Index for Connecticut created by Open Communities Alliance and contributors, and is based on 11 variables from similar public data sources, as described below.

To download the data, go here. For more on the use of Opportunity Mapping in Connecticut, see the websites for Open Communities Alliance.

 

Geography

Neighborhoods: We use Census tracts as neighborhood proxies.  Census tracts are small, relatively permanent geographic subdivisions of a county or equivalent entity; they generally have a population between 1,200 and 8,000 people, with an optimum size of 4,000. We use Census tract boundaries from the 2010 decennial census. There are 824 Census tracts in the state with opportunity index scores.  All 11 indicators use Census tract data and the comprehensive index is computed for tracts as well.

 

Data Sources and Indicators

The comprehensive opportunity index is a combination of 11 separate indicators in three sub-areas:  (1) Educational, (2) Economic, (3) Neighborhood/Housing quality. 

 

Educational indicators:

  1. Share of population 25 to 64 years with some college, BA degree, or higher. Data by Census tract are from the 2013-2017 American Community Survey, Table B23006: Educational Attainment by Employment Status for the Population 25 to 64 Years.

  2. Connecticut State Department of Education Next Generation Accountability scores: Connecticut’s Next Generation Accountability System is a broad set of 12 indicators that help tell the story of how well a school or district is preparing its students for success in college, careers and life. The system moves beyond test scores and graduation rates and instead provides a more holistic, multifactor perspective of district and school performance and incorporates student growth over time. It was developed through extensive consultation with district and school leaders, Connecticut educators, state and national experts, CSDE staff, and many others. The system was conceived and developed under ESEA Flexibility and approved by the U.S. Department of Education (USED) on August 6, 2015. It was later included as part of Connecticut’s state plan under the Every Student Succeeds Act (ESSA). We use CSDE’s overall accountability index scores for districts in 2017-18, and which reflect a combination of the 12 next generation factors.  District-level scores are matched to Census tracts using school district-to-tract crosswalk files.  If Census tracts overlap into multiple districts, a summary tract index score is computed by weighting the district index score by the share of tract population in the respective districts.

 

Economic indicators:

  1. Percent in the civilian labor force but unemployed. Data by Census tract are from the 2013-2017 American Community Survey, Table B23025: Employment Status for the Population 16 Years and Over. To ensure that large values on each indicator denotes high opportunity, we compute the inverse percentage (i.e. 1 – rate). 

  2. Percent Change, Annual Average Employment, 2014-17. This indicator relies on data from the Quarterly Census of Earnings and Wages (QCEW) series from the Department of Labor. As noted on their website, the QCEW program serves as a near census of employment and wage information. The program produces a comprehensive tabulation of employment and wage information for workers covered by Connecticut Unemployment Insurance (UI) laws and Federal workers covered by the Unemployment Compensation for Federal Employees (UCFE) program.  We employ data at the town (i.e. county subdivision) level, and match these data to Census tracts using tract-county subdivision crosswalks.  If Census tracts overlap into multiple county subdivisions, a summary tract value is computed by weighting the county subdivision value by the share of tract population in the respective county subdivisions.  Once each year’s employment data are matched to tracts, the percentage change measure is computed.

  3. Employment Access: This indicator uses data drawn from the Location Affordability Index(LAI), version 3.  The Employment Access Index is determined using a gravity model which considers both the quantity of and distance to all employment destinations, relative to any given Census tract. Using an inverse-square law, an employment index is calculated by summing the total number of jobs divided by the square of the distance to those jobs. This quantity allows for the examination of both the existence of jobs and the accessibility of these jobs for a given Census tract. Because a gravity model enables consideration of jobs both directly in and adjacent to a given Census tract, the employment access index gives a better measure of job opportunity, and thus a better understanding of job access than a simple employment density measure. This index also serves as a surrogate for access to economic activity.

  4. Job Diversity: This indicator uses data drawn from the Location Affordability Index(LAI), version 1.  The jobs diversity index looks at the correlation between 20 major job sectors - areas with higher concentration in a few sectors are reported as having lower diversity.  See the LAI documentation for more details on the computation of the jobs diversity index.  Block-group level LAI data were summarized to the tract level by computing a weighted average of the index value, weighted by the number of households in each tract.

  5. Share of households receiving public assistance income or food stamps in past 12 months. Data by Census tract are from the 2013-2017 American Community Survey, Table B19058: Public Assistance Income or Food Stamps/SNAP in the Past 12 Months For Households.  To ensure that large values on each indicator denotes high opportunity, we compute the inverse percentage (i.e. 1 – rate). 

 

Neighborhood / housing quality indicators:

  1. Share of population in owner-occupied housing. Data by Census tract are from the 2013-2017 American Community Survey, Table B25008: Total Population in Occupied Housing Units By Tenure.

  2. Crime: Crime rates are reported by local authorities to the Uniform Crime Reports database. Rates are calculated as the number of incidents in a town divided by the current population of the town. For this index, the 2015 crime rates are used as they are the most recent and readily-available data for the state.  The indicator includes murder, rape, robbery, Aggravated Assault, Burglary, Larceny, Motor Vehicle Theft, and Arson.  These crime rates are matched onto Census tracts using a tract-county subdivision crosswalk.  If Census tracts overlap into multiple county subdivisions, a summary tract value is computed by weighting the county subdivision value by the share of tract population in the respective county subdivisions.

  3. Percent of housing units that are vacant. Data by Census tract are from the 2013-2017 American Community Survey, Table B25002: Occupancy Status. To ensure that large values on each indicator denotes high opportunity, we compute the inverse percentage (i.e. 1 – rate). 

  4. Percent of households with income in the past 12 months below the poverty level. Data by Census tract are from the 2013-2017 American Community Survey, Table B17017: Poverty Status in the Past 12 Months by Household Type By Age of Householder.  To ensure that large values on each indicator denotes high opportunity, we compute the inverse percentage (i.e. 1 – rate). 

 

Methodology

The Opportunity Index is computed by averaging the standardized component indicators.  In other words, the final opportunity index score for each census tract is based on the average z-score for all indicators. 

The opportunity levels (very low, low, moderate, high, very high) are based on quintiles of tract opportunity index scores. For instance, “very high” opportunity tracts are those with opportunity index scores in the top quintile. Conversely, “very low” opportunity tracts are those with index scores in the bottom quintile.   If a Census tract is missing data for any of the 11 indicators that make up the composite opportunity index, the opportunity index is computed using the available data.  There are only four Census tracts for which this is the case: 09005293100 (Colebrook), 09003500700 (Hartford), 09013538100 (Somers), and 09005265100 (Warren).

The Hartford and Somers Census tracts do not have any housing units, and therefore only 4 of the variables have a value. Because these tracts do not have housing units, we are comfortable with approximating an opportunity level for these tracts because functionally it will not be relevant to decisions for Connecticut residents. Colebrook and Warren Census tracts, on the other hand, lack large enough job markets for there to be data with regard to job changes, so the opportunity index for these Census tracts was calculated from 10 of the 11 variables.

 

A Note on Margins of Error

Six of our indicators employ data from the U.S. Census American Community Survey (ACS).  ACS data are survey data and are therefore subject to sampling error.  Sampling error occurs when a sample is taken instead of observing the entire population, and is the difference between the sample estimate and the true but unobserved population parameter.  The Census Bureau provides margins of error (MOEs) for each census enumeration area (in our case Census tracts) for each estimate in ACS tables.  These margins of error quantify the magnitudes of sampling error, and should be used alongside the estimates in sound statistical analyses to evaluate the reliability of the estimates.  However, the magnitude of margins of error is relative to the size of the estimates, such that estimates with larger values also have larger margins of error.  As such, larger MOEs do not necessarily imply less reliable estimates.  Coefficients of variation (CVs) are therefore preferable measures to determine estimate reliability insofar as they are independent of the scale of the estimates.  The CV is equal to the standard error divided by the estimate, and as such indicates the relative amount of sampling error associated with the estimate.  We compute CVs for the six indicators that use ACS data: (i) Percent in the civilian labor force but unemployed; (ii) Share of households receiving public assistance income or food stamps in past 12 months; (iii) Share of population 25 to 64 years with some college, BA degree, or higher; (iv) Share of population in owner-occupied housing; (v) Percent of housing units that are vacant; and (vi) Percent of households with income in the past 12 months below the poverty level. 

The process for computing these CVs is straightforward.  First, we used the ACS handbook to derive margins of error for the proportions that we generated using ACS estimates.  These derived margins of error employ the 90% MOEs for estimates provided by Census in the ACS.  For instance, our vacancy indicator is based on two ACS estimates (one for the numerator and a second for the denominator), and we use the two corresponding Census-provided 90% MOEs for these two estimates to generate the derived MOE for the share of vacant housing units.  We then use these derived MOEs to compute CVs according to the formula below.  CVs are provided in percentage form by multiplying by 100.  These CVs are available in this file.  Analysts can use these CVs to evaluate the reliability of indicators that are based on survey data.

 

CV_Picture1.png

 

 

 

 

  • Open Communities Alliance
  • 75 Charter Oak Avenue
  • Suite 1-210
  • Hartford, CT 06106
  • Phone: 860-610-6040