We have gone through this before—how President Trump, the White House and their media allies have cited national polls with extremely small U.S. Latino samples to claim that the nations’s largest ethnic group is fully behind the current administration.
The latest example comes from the White House’s official Twitter profile, which on Tuesday tweeted the following:
Democrat lawmakers have claimed that including a citizenship question on the next U.S. census will discourage Hispanic voters from participating.
A majority of Hispanics say they favor a citizenship question.https://t.co/Ntd4VeJcIa
— The White House (@WhiteHouse) July 9, 2019
Sounds pretty definitive, right?
As suspected when it comes to national polls that uses Latinos (and other racial, ethnic and demographic) as subgroups, all one needs to do is check the poll’s crosstabs to find out exactly how large of a sample was uses for each subgroup. In this June poll, 2,182 registered voters participated. The weighted results included several characteristics, although the poll missed one big question to ask U.S. Latino voters: What language is spoken in your home? Strike one for not creating a poll that fully and accurately represents the entire U.S. Latino community.
Of those 2,182 participants, 13% of those surveyed said they were Hispanic. That means that the poll had 284 people who said they identified as Hispanic. That’s not a large sample, and it is very likely not a representative sample.
How can we conclude that (outside of the language issue)? Here’s another way: by looking at state breakdowns from Harvard/Harris (pages 6-9). If you take the top 10 U.S. states with the largest Latino populations, for example, and compare them to the state breakdowns that Harvard-Harris provided, it’s clear that this poll does not even begin to accurately reflect the community overall. In fact, based on the following quick analysis, the poll is underrepresented in places like New Jersey, Illinois, New Mexico, Colorado and Georgia (yes, Georgia).
|2016 Latino Population (US Census)
How underrepresented are these states? Think about this: the poll asked 3 Latinos in New Hampshire and 3 in Nebraska, but only asked 3 Latinos in New Jersey and 3 in Colorado? Or just 1 Latino in Massachusetts?
Finally, if we take the 55% of U.S. Latino support for a citizenship question as actual data (and we don’t, but let’s say we do), it would mean that the White House is basing its tweet on the support of 156 Latinos.
156 Latinos who are very likely (and this is super important) not an accurate and representative sample of the U.S. Latino population.
But let’s assume they are (even though they are not). At best, the current Harvard/Harris poll would have a margin of error of +/- 5.82%, which is not as a high as the last time we did a Harvard/Harris Latino respondent analysis, but it’s still pretty high.
Still, no matter now many strikes you give Harvard/Harris for continuously missing the mark on accurately polling U.S. Latinos, the damage has already been done. The White House and its media allies will continue to take advantage of limited and inaccurate data to create a narrative that just isn’t reality. That’s how propaganda works, and there are enough voices out there who will chose to believe whatever the Trump and the White House tell them.
Once can slice all the data one wants, but it’s too late.
Or is it?
Maybe people should start by refreshing their memory and citing more accurate polls of U.S. Latinos that state this (from 2018): 93 percent of U.S. Latinos believe “it is important for the Census Bureau to implement a complete and accurate count of the entire Latino population and that nobody is left out.”