Thursday, August 2, 2012

Voter ID impact on minority voters in Philadelphia

Voter ID impact on minority voters in Philadelphia

Tamara Manik-Perlman, Project Manager & Spatial Data Analyst at Azavea, a did the Geographic Information Systems work and statistical analysis at the heart of this post. Details of her work, along with maps and charts, are available here.

Exec summary

Pennsylvania’s new Voter ID law affects African and Hispanic voters at a much higher rate than whites in Philadelphia, according to analysis involving U.S. Census data, voter registration records and Voter ID provided by Pennsylvania's Secretary of State.

Statistical correlation shows a strong relationship between the racial/ethnic makeup of voting precincts and the percentage of voters who lack state ID required to vote or whose state IDs have expired and are not usable.

A voter who lives in the city’s most heavily African American voting precincts is 85% more likely to lack a valid ID than a voter who lives in a predominantly white precinct. Voters who live in heavily Hispanic areas are 108 percent more likely to lack valid state ID than those from predominately white areas. Heavily Asian neighborhoods show the same pattern. 

If new state IDs are not produced for tens of thousands of Philadelphia residents, the analysis indicates, Voter ID will significantly reduce the vote of African-Americans, Hispanics and Asians relative to the city's overall population.

Philadelphia Ward-divisions (Precincts), showing percentage of voters with either no ID or expired ID. Blue means lower percentage of voters with problems; yellow and orange means a higher percentage with problems.

Map legend


The Pennsylvania Secretary of State’s office, after previously estimating that about 1 percent of Pennsylvania voters would have to get new photo IDs to satisfy the requirements of the law, earlier this summer began an attempt to identify these voters. 

The state compared its database of voters with the Pennsylvania Department of Transportation’s database of driver’s licenses and non-driver ID cards.  The exact method of comparison has not been disclosed, but in early July, the Pennsylvania Secretary of State sent to county voter registrars a list of 758,000 registered voters (186,000 in Philadelphia) who could not be located in the PennDOT database. We refer to this as the “no id” list.

In late July, county registrars were sent another list of 576,000 voters (176,000 in Philadelphia) who had been matched in the PennDOT database but whose licenses or IDs had expired before November 2011 and would therefore be unusable for voting in November 2012. This we refer to as the “expired” list. 

We obtained the two lists from Philadelphia Commissioner Stephanie Singer’s office and compared the two based on the voter identification number, determining that there is no overlap between the two lists. In all, 362,329, or 35 percent of Philadelphia’s registered voters, are either on the “no id” or “expired id” list received by Commissioner Singer’s office. Of the 868,000 Philadelphia voters regarded as “active,” having voted in the last 4 years, 282,609, or 32.5 percent is either on the “no id” or “expired” list.

Limitations of the data and caveats

While the “no id” and “expired id” data represents the best available information on the likely impact of Pennsylvania’s Voter ID law – indeed it may be the best information available to date from any state that has enacted such laws – there are limitations.

The “no id” list, as The Philadelphia Inquirer reported July 29, contains voters who in fact have IDs but weren’t found on the PennDOT list because of name discrepancies between the two databases or simple spelling errors. One example is voters who registered under their maiden name, were subsequently married, and then obtained a driver’s license under their married name. At least some of these voters presumably would be permitted to vote if they could provide supplemental identification to prove they are in fact the same person, but it is not clear how many could do so.

In fact, it is impossible to know how many on the “no id” list will be unable to vote without obtaining state IDs between now and November, though it is probably fair to say it will greatly exceed the 1 percent impact predicted by Gov. Corbett and Secretary of State Aichele at the time Voter ID was enacted. 

Even if one assumes that half of the “no id” voters in fact turn out to have IDs that can be used to vote, that would still leave more than 200,000 Philadelphians who are active voters who have either no ID or expired ID.

The expired id list, which was released later in July to county voter registrars, is probably a “harder” number because it represents positive matches, not the absence of a match. Assuming the Secretary of State did not make any mistakes in the analysis, the “expired” list represents voters who were found on the PennDOT database and whose licenses or IDs had expired before November 2011 and had not been renewed at the time the data snapshot was taken.

Measuring impact by race/ethnicity

In the end our report is not trying to answer the question of total magnitude; our focus is on race and ethnicity and disproportionate impact. Whether the percentage of Philadelphia voters affected is 35 percent or 20 percent, or some smaller number, the purpose of this analysis is to make a first attempt to determine whether the law will affect minority voting rights in Philadelphia.

The federal Voting Rights Act of 1965, renewed by Congress in 2006, prohibits states from imposing any "voting qualification or prerequisite to voting, or standard, practice, or procedure ... to deny or abridge the right of any citizen of the United States to vote on account of race or color." The U.S. Department of Justice has launched an inquiry into Pennsylvania’s Voter ID law to determine whether it affects minority voters disproportionately.

Philadelphia presents an excellent data set with which to explore this question. While the city overall is racially diverse, many parts of the city remain ethnically homogeneous and therefore provide us with a way to compare Voter ID impact by ethnicity. 

Moreover, Pennsylvania’s data could have national implications, since Pennsylvania’s law is similar in language to several other states that have enacted Voter ID in 2011 and 2012. If the law negatively affects the minority vote in Philadelphia to the degree that we have found, the same pattern may be found in other Voter ID states.

This represents only an initial look at the data, using very basic techniques that can be easily duplicated. We hope academic researchers, news media, voting-rights organizations and others can use what we’ve done and verify and amplify these findings.  This paper will fully disclose the techniques used, and we will share the raw data – which is public information – with interested parties.

Data used:
  •  Philadelphia’s voter registration electronic file, a public document. The file contains 1.02 million voter records, of whom 868,000 are considered active voters because they have cast ballots in the last four years.
  •   The “no id” file, an electronic file from Pennsylvania Secretary of State Aichele's office containing 186,560 Philadelphia voters who were not found in the Pennsylvania Department of Transportation database of driver’s licenses and non-driver state IDs.  We refer to this is the “no id” file.
  •  A second file provided by the Secretary of State in mid-July, containing an additional 175,769  Philadelphia voters who do have PennDOT IDs, but whose IDs expired before November 2011 and would therefore be unusable in November 2012 elections under the Pennsylvania law.)
  • U.S. Census data for 2010 – 100 percent counts for Philadelphia, broken down by Census block.
  • A Geographic Information Systems boundary file of 1,687 Philadelphia ward divisions – otherwise known as precincts.


“No ID” and “expired ID” files were joined to the Philadelphia voter registration file using the state voter identification number present in all three files.  This is a simple one-field join for a SQL database and can also be accomplished by some statistical software packages.

With the data combined, crosstabs could be run counting the total number of voters, and the number on the “no id” or “expired” lists, for each ward division in the city.

Meanwhile, using GIS software, Azavea merged census block data for population and ethnicity into the boundary file for Philadelphia ward divisions. The result is highly accurate breakdowns of voting population by race/ethnicity for voting-age population in the 1,687 ward divisions. 

Using the divisions as the unit of analysis, Pearson’s R correlations were run in Microsoft Excel. There would be many other statistical tools that could be used, including multiple linear regression. But we chose correlations for simplicity and also because it can be duplicated on desktop software that many people already own. 

Results 1: Correlations between ethnic composition of ward divisions and percentage of voters with possible ID problems.

Using 1,687 ward divisions as the unit of analysis, a strongly positive correlation was found between the percentage of voters with potential ID problems and the percentage of African-American population voting-age population. Which means: the higher the African-American percentage in the ward division, the higher the percentage of voters on the “no id” or “expired” id lists.

Here are the correlation coefficients obtained:
·         African-American: + 0.50
·         White: - 0.65
·         Hispanic: + 0.26

 Here's the plot of Philadelphia precincts based on the percentage of Black voters. As the proportion of Black voters in a preinct goes up, so does the percentage of voters who lacking state ID cards or have expired state IDs.

 For white voting voting-age population. A strongly negative correlation was found, meaning the higher the percentage of white population in the district, the fewer voters with ID problems.

Here's the plot based on the percentage of white voters in Philadelphia precincts. As the proportion of white voters in a precinct goes down (the bottom axis), so goes the percentage of voters with ID problems.

For Hispanic/Latino population, a weaker correlation was found, mainly because the Hispanic percentage in the city is lower, but the pattern is the same.

However, we believe this result does not mean Hispanics are less affected by Voter ID than African-Americans. When the analysis is re-run only on ward divisions where the percent Latino is 10% or greater, the correlation is +0.65-- stronger even than for African-Americans.  Similarly, when the procedure is run on divisions with > 10% Asian voters, a moderately positive +0.22 correlation is found, which means the higher the Asian population, the more voters with ID issues.

Results 2: Direct comparisons of White, African American, Hispanic and Asian areas.
Another way to look at the data, and one which may be more understandable to lay readers is to compare ward divisions that have the highest percentages of one race or another. 

Out of 1,685 ward divisions in the city (2 were excluded because they have almost no population) we selected the top 10 percent (169) of divisions by ethnic concentration.

The top 169 most heavily white divisions, according to the Census, contain a voting age population approximately 93 percent white. The percentage of voters who lack state ID or have expired ID in these divisions (“no id” or “expired”) is 21.0 percent, which is our baseline for comparison.

Next, we compare to the 169 divisions with the highest percentage of African-American voting-age population. As a group, these areas are 96 percent black. The percentage of voter id problems for this group is 38.9 percent.

To summarize, a voter who lives in these predominantly African American neighborhoods is 85% more likely to lack a state ID than a voter who lives in one of the most predominantly white precincts.

Hispanic impact

We selected the 169 divisions with the highest percentage of Hispanic voting age population according to the census. As a group, these divisions are 54 percent Hispanic. The percentage of voter id problems for this group is 43.6 percent.

So based on this comparison, A voter who lives in these heavily Latino neighborhoods is 108% more likely to lack the required ID to vote than someone who lives in Philadelphia’s predominantly white precincts.

Asian impact is more difficult to measure in this way because Philadelphia’s overall Asian population is smaller and there are fewer neighborhoods that are as highly concentrated.  

We selected the 169 divisions with the highest percentage of Asian voting age population according to the census. As a group, these divisions are 16 percent Asian. The percentage of voter id problems for this group is 28.3 percent. A voter who lives in these neighborhoods with a large Asian presence is 35% more likely to lack the required ID to vote than someone who lives in Philadelphia’s predominantly white precincts.

If we select a smaller group of Asian precincts with higher concentrations of Asian voting age population, the same pattern emerges that is seen with other ethnic minorities. 

We selected 11 ward divisions where the Asian voting age population was over 40 percent. These divisions have a total voting age population of 13,720, and contain 4,978 active voters. For this group of precincts, the percentage of active voters with ID problems indicated is 38.2 percent. So a person living in one of these districts is 82 percent more likely to have problems than in the predominantly white neighborhoods.

Further analysis

We intentionally kept the analysis simple and understandable, with a minimum of adjustments to the data. For example, correlations would be slightly stronger if college students and or college precincts, where there are large numbers of 2008 voters who have since moved out of Pennsylvania, were isolated out of the analysis.

We did not explore more advanced statistics, such as multiple regression, or more elaborate geo-statistical techniques that might identify clusters in the city.

The question we set out to answer was a simple one -- does Voter ID affect minorities more than the average resident? And the answer, though not surprising, is a clear yes.

No comments:

Post a Comment