Poverty in the United Way Service Area

Transcription

1 Poverty in the United Way Service Area Year 2 Update 2012 The Institute for Urban Policy Research At The University of Texas at Dallas

2

3 Poverty in the United Way Service Area Year 2 Update 2012 Introduction The United Way of Metropolitan Dallas established bold community-wide goals in the areas of Income, Health, and Education. In the area of Income, the United Way established a goal of helping 250,000 fewer persons exit poverty than would otherwise have been expected without United Way involvement. This report details the state of poverty in the United Way service area in 2012, and compares this to the condition set at baseline in It compares changes in poverty since 2010 against what would have been expected had the trend at baseline continued uninterrupted. The report is supplemented by an additional report that further details the prevalence of poverty across multiple demographic groups. The process of developing these estimates and projections is technically complex. First, census boundaries do not precisely align with the United Way s service area. Furthermore, these boundaries were significantly changed in 2012, so a discussion of the precise methods used to approximate the geography is presented in Appendix 1. Second, because the estimates and projections below are derived from data sources that represent only a sample of the population, there is some margin of error around them. Appendix 1 contains a more robust discussion of the process used to generate the margins of error, an indicator of the reliability of each estimate, and an indicator of statistically significant changes from baseline for each indicator. Poverty in the United Way Service Area As has been documented in prior reports, the Institute s analysis of the U.S. Census Bureau s American Community Survey (ACS) indicated there were an estimated 563,874 persons living in poverty in the United Way of Metropolitan Dallas service area (Dallas, Collin, Rockwall, and southern Denton counties) in That report also detailed the Institute s projection of approximately 967,515 persons living in poverty in the service area by Using the same estimation methodology as previous years, there were an estimated 582,868 persons living in poverty in the service area in 2012, more than 54,000 fewer than the 636,879 that would have been anticipated with no additional intervention. Figure 1 depicts these data, with green dots representing the number estimated to be in poverty each year through The solid red line depicts the projected change through 2020 if the pattern established through baseline continues. Table 1 presents the year by year numbers estimated from the American Community Survey, as well as projected numbers for the following years through P a g e 3

4 Figure 1. Number of Persons Living Below Poverty, Table 1. Number of Persons Living Below Poverty, Number of Poor Persons Number of Poor Persons Expected With No Change to Current Trends , , , , , , , , , , , , , , , , ,515 1 Source: Institute for Urban Policy Research analysis of Public Use Microdata Sample (PUMS) data from the American Community Survey 1 year estimates 2005, 2006, 2007, 2008, 2009, and 2010, P a g e 4

5 Table 2 presents the numbers estimated to be in poverty, along with the percent of population. For reference, Table 4 presents the federal poverty threshold for the year This is the guideline that was employed by the Census Bureau in assigning poverty status to household members. Table 2. Number and Percent of Persons by Poverty Status, Poverty Status At or Above Poverty 2,910,063 3,059,956 3,095,304 3,165,426 3,189,065 3,113,406 3,151,335 3,239, Below Poverty 452, , , , , , , , Table 3. Federal Poverty Threshold by Size of Family Unit and Number of Children under Age 18, 2012 Household Income Number of Related Children Under 18 Years Size of Family Unit or More One Person Under 65 11, and Up 11,011 Two People HH Under 65 15,374 15,825 HH 65 and Up 13,878 15,765 Three or More People Three people 17,959 18,480 18,498 Four people 23,681 24,069 23,283 23,364 Five people 28,558 28,974 28,087 27,400 26,981 Six people 32,847 32,978 32,298 31,647 30,678 30,104 Seven people 37,795 38,031 37,217 36,651 35,594 34,362 33,009 Eight people 42,271 42,644 41,876 41,204 40,249 39,038 37,777 37,457 Nine people or more 50,849 51,095 50,416 49,845 48,908 47,620 46,454 46,165 44,387 Considerations in Assessing Changes in Estimates While the margins of error and statistical significance are discussed more fully in Appendix 1, the reader should be cautioned to consider the following information when assessing year to year changes. The American Community Survey (ACS) draws on responses from a small subset of the population, and the Public Use Microdata Sample (PUMS) data used to prepare this report draws on a still smaller subset of that data. Thus all estimates presented for 2012 (and for prior years) are accompanied by a margin of error in Appendix 1. The narrower the margin, the more reliable the estimate. This, in essence, is the driver behind the symbols used to mark the estimates in Appendix 1 as high, moderate, or low reliability. The issue is further exacerbated when comparing estimates from two or more years. Quite simply, it could be noise that s responsible for a shift in the percent or number of persons in poverty, particularly in smaller subsets of the population. To that end, Appendix 1 also presents indicators of statistical significance for each change figure. To guard against unwarranted conclusions, the report suppresses indications of statistically significant changes when one or both of the estimates being compared were marked as low reliability. P a g e 5

6 Appendix 1: Reliability of the Estimates A Summary of the Data Source The estimates presented in this report were derived from the Public Use Microdata Sample (PUMS), which is a representative sample of individual records drawn from the American Community Survey (ACS). They represent a roughly 1% sample of the nation's households, and all of the persons in each of the sampled households. Because the estimates are drawn from a sample that is itself drawn from a sample of the population, significant care must be taken in estimating the reliability for each percentage or total computed. Considerations of geography and sampling strategy are outlined below. Geographic Implications The United Way of Metropolitan Dallas serves Dallas, Collin, Rockwall, and southern Denton counties. The data in PUMS are made available at a unit of geography known as the Public Use Microdata Area, or PUMA. PUMAs are sufficiently large so as to ensure confidentiality of census respondent information. In counties like Dallas, the sheer number of persons allowed there to be 15 PUMAs in 2000, increasing to 22 in When counties have smaller populations, they re often combined to create one PUMA. In the 2000 delineations, Rockwall and Kaufman counties were combined. For the 2010 delineations, Rockwall was combined with Hunt. This has implications for the analysis of ACS PUMS data for the United Way service area. Figure 2. Alignment of United Way Service Area and PUMAs, 2010 Figure 2 illustrates the alignment between the United Way service area, outlined in blue, and the 2000 Census PUMAs that were aggregated to comprise the approximations used in the 2010 and 2011 reports, outlined in red, and the 2010 Census PUMAs that were aggregated to comprise the P a g e 6

7 approximations used in the 2012 report, outlined in yellow 2. While the counties of Dallas, Rockwall, and Collin are completely contained, the approximation area includes portions of Denton County that fall beyond the service area. In the 2010 and 2011 reports, it also included all of Kaufman County, while the 2012 report, which used the 2012 PUMAs, dropped Kaufman County and added Hunt County. The inclusion of Kaufman and Hunt counties over different years has minimal implications for relative prevalence (e.g., percentages), but Kaufman County and Hunt County does add approximately 80,000 to 100,000 persons into the formula. However, with an aggregate population in the approximation area of almost 3.7 million, the influence of 100,000 persons is negligible. For the overall measure of percent and number of persons in poverty, the 2012 estimates (based on the newly drawn PUMAs) were adjusted back to the 2010 and 2011 estimate boundaries. Dallas and Collin counties were included in their entireties, with no impact on estimations. Denton County s 2010 PUMAs were adjusted back to 2000 PUMAs by using the geographic correspondence service for population base counts hosted by MoCeDA. The aggregate counts were adjusted downward by removing Hunt County s poor and non-poor populations, and adjusted upward by adding Kaufmann County s poor and non-poor populations. All other estimates use the new geographies. Reliability of the Estimates Each record included in the PUMS data is weighted to reflect the probability of that record having been selected into the sample. This weighting is a method of controlling for variations in the sampling procedure designed to ensure representation across various dimensions. An additional set of 80 weights is generated by the Census Bureau for each record using a method known as Successive Difference Replicates (SDR) Weighting. To assess the reliability of the estimates prepared above, they are actually reproduced 80 times using each of the different SDR weights. The standard error of the estimate is generated then from the 80 differently weighted versions to produce a standard error that recognizes the sample from a sample issue peculiar to PUMS data. The tables that follow provide, for each percentage and total provided in the report, the standard error for the estimate that was produced using the SDR methodology. The standard error can be thought of as one indicator of the reliability of the estimate, in that the larger the standard error the less reliable the estimate is. The standard error is then used in the computation of a 95% confidence interval around the original estimate. The lower and upper bounds of the confidence interval are reported in the table as well. Finally, using the coefficient of variation (CV) as a guide, we provide an indicator of the reliability of each estimate. 3 When the coefficient of variation falls at or below 12%, the estimate is thought to be of high reliability. When the coefficient of variation falls above 40%, the estimate is thought to be of low reliability. When the coefficient of variation falls in between, the estimate is said to be of medium reliability. The level of reliability is indicated in the tables below by a green circle for high reliability, a yellow triangle for medium reliability, and a red diamond for low reliability. 2 The following 2000 PUMAs were aggregated to comprise the United Way Service Area: 2000, 2101, 2102, 2103, 2104, 2201, 2301, 2302, 2303, 2304, 2305, 2306, 2307, 2308, 2309, 2310, 2311, 2312, 2313, 2314, and For the 2012 report, the following 2010 PUMAs were aggregated: 900, 1901, 1902, 1903, 1904, 1905, 1906, 1907, 2001, 2002, 2003, 2301, 2302, 2303, 2304, 2305, 2306, 2307, 2308, 2309, 2310, 2311, 2312, 2313, 2314, 2315, 2316, 2317, 2318, 2319, 2320, 2321, and For a complete discussion of the methodology, see National Research Council, Using the American Community Survey: Benefits and Challenges (Washington, D.C.: The National Academies Press, 2007). P a g e 7

8 Table 4. Reliability Indicators for Percent of Persons in Poverty, Year Percent Standard Error Lower Bound Upper Bound Reliability Table 5. Reliability Indicators for Number of Persons in Poverty, Year TotalStandard Error Lower Bound Upper Bound Reliability ,178 18, , , ,562 15, , , ,950 16, , , ,388 17, , , ,508 17, , , ,874 15, , , ,159 18, , , ,868 14, , , Statistical Significance of the Changes Recall that the American Community Survey is a survey of a sample of the population. The Public Use Micro sample Data used to produce the estimates contained herein are a sample drawn from the ACS sample. For that reason, the estimates that were presented in this report were accompanied by a margin of error computed at the 90% level. Comparing each year to the previous year presents further complications. When we compute the change in the number of percent of persons living in poverty, we must treat that difference to the same cautious interpretation. The formula for the standard error of the difference between two years estimates is simply derived by taking the square root of sum of each year s squared standard deviation. In that vein, we present below tables detailing the margin of error and statistical significance of changes over time. For each change in either the number or percentage of persons or households, we provide the standard error of the difference, as well as the lower and upper bounds of the 90% confidence interval. In addition, we graphically present the statistical significance of the change at the 0.10 α level. We depict statistically significant upward progress (fewer number or percent poor) with a green upward pointing arrow. Non-statistically significant change is presented with a yellow dash, while statistically significant negative change (higher numbers or percent poor) is presented with a red downward pointing arrow. P a g e 8

9 Table 6. Significance of Annual Change in Percent of Persons in Poverty, Year Percent Standard Error Lower Bound Upper Bound Significant Change Table 7. Significance of Change from Base Year in Percent of Persons in Poverty, Year Percent Standard Error Lower Bound Upper Bound Significant Change Table 8. Significance of Annual Change in Number of Persons in Poverty, Year Total Standard Error Lower Bound Upper Bound ,285 24,142.0 (9,428.6) 69, ,291 23,500.7 (49,949.6) 27,367.6 Significant Change Table 9. Significance of Change from Base Year in Number of Persons in Poverty, Year Total Standard Error Lower Bound Upper Bound ,285 24,142.0 (9,428.6) 69, ,994 21,420.8 (16,243.3) 54,231.3 Significant Change P a g e 9

10 800 West Campbell Road, GR20 Richardson, Texas (972) voice (972) facsimile Visit us on the Web!