Emails, responses to PolitiFact Texas, Lynda Tran, partner, 270 Strategies, Andrew Claster, data contractor to 270 Partners, March 1 and March 20-21, 2013

10:31 a.m.

Hey there,

Lynda Tran here - I'm partner at 270 Strategies and working on Battleground Texas.

Have a lot of information for you here. To run counts on the number of Latino citizens aged 18+ in Texas, use the US Census American Community Survey Public Use Microdata Set. That information can be obtained using this link:

For Texas voter file data, contact the Elections Division at the Texas Board of Elections:

If you are interested in self-reported data, you can use the Voting and Registration in the Election of November 2008 from the US Census:

According to this survey, 54.3% of Texas Latino citizens aged 18+ reported that they are registered to vote and 37.8% reported that they cast a ballot.


4:47 pm

March 20, 2013

Colbert Report Feb 26

Minute 4:14

Jeremy Bird: “In 2008, for example, only 54% of Latinos in Texas were registered to vote and only 35% actually turned out”

The 54% number comes from US Census Voting and Registration in the Election of November 2008

Table 4b

Cell G 543

And here is how we calculated the 35% number:

Denominator = number of TX Latino citizens aged 18+ in 2008

Source: Census Public Use Microdata Set 1-year estimate 2008

1340097 3,812,343

Numerator = number of Texans casting ballots in Nov 2008 who are Latino

Source: DNC voter file, acquired from TX Secretary of State Elections Division, with commercial ethnicity model applied to determine ethnicity based on name and Census info


Numerator / denominator = 35%

Note that the Census survey reports that 38% of TX Latino citizens aged 18+ cast ballots in Nov 2008

Table 4b

Cell L 543

(W. Gardner Selby, PolitiFact Texas)

5:58 pm

March 20, 2013

So, you pulled this figure from PUMS--3,812,343—though the CPS survey estimate, also a single-year result for 2008, is that there were nearly 4.5 million voting-age Latino citizens in 2008. Correct?


If so, this means that you took a high-end figure as your denominator to come up with the 54 percent figure aired in the interview but a Texas Latino VAP citizen figure about 700,000 lower to get the other percentage. Why isn’t that mixing apples and oranges and statistically and/or logically questionable?





(Claster reply)

10:05 am

March 21, 2013

The numbers Jeremy used on the show were 54% and 35%.  The 35% number is close to the 38% number.  Obviously Census produced two different estimates in 2008: 3.81 million and 4.49 million.  If the numerator is 1.34 million (from the voter file) then if we use PUMS for the denominator, we get 35%.  If we use Census CPS for the denominator, then we get 30%.  The lower number actually helps make Jeremy's case stronger - but 35% is a central estimate, while 30% and 38% are more extreme.  Instead of saying '35%', Jeremy might have said 'between 30% and 38%, depending on sampling error and whether one relies on the Census CPS survey or the Census PUMS 1-year estimates combined with the voter file from the Texas Secretary of State'.  But I imagine you can understand why he chose not to say that on the Colbert show.