Unfortunately, you can’t look at the 2010 and 2011 values for Travis County and conclude that poverty went down. SAIPE is a model, not a direct count of poverty (like the decennial census), and if you want to make year-to-year comparisons, the 90% confidence interval (the right-most column provided for SAIPE estimates) has to be taken into account. For Travis County, the poverty rate estimate is
2010: 17.0 to 19.2%
2011: 17.8 to 19.8%
The overlap between the two estimates’ ranges means you can’t conclude that poverty went down. I’m pretty sure a valid “change in poverty” comparison would require a three-year average of the Travis county estimates, but you can check with the Census Bureau to make sure. (Two-year average might be sufficient.) In other words:
2008-2010: 16.4% poverty
2009-2011: 17.6% poverty (an increase)
Feb. 25, 2013
Here’s what the ACS (Table B17001) says for Travis County:
2011: 192,436 residents with income below poverty level, plus or minus 13,557 (ACS, like SAIPE, is a survey, and this is their margin of error).
2010: 194,156, plus or minus 13,427
2009: 163,630, plus or minus 11,214
So, to see if there are statistically significant changes among any of those years, you’d compare the ranges:
2011: 178,879 to 205,993
2010: 180,729 to 207,583
2009: 152,416 to 174,844
2010 and 2011 overlap enough (as they did in the SAIPE rates) that you can’t conclude there’s been any statistically significant change in the number of poor people living in Travis County. However, there’s no overlap at all with the 2009 range; that means the number of poor people HAS gone up in either 2010 or 2011, compared to 2009. You see something similar if you look at the ACS poverty rate estimates (with margins of error), Table DP03:
2010: SAIPE: 17.8 to 19.8% ACS: 17.9% to 20.5%
2011: SAIPE: 17.0 to 19.2% ACS: 17.2% to 19.8%
2009: ACS: 15.1 to 17.3% (statistically sig. lower than 2010, but not 2011 [barely – 17.3 vs 17.2])
Also: “Low income” and “poverty” are not interchangeable terms; “low income” means two times the poverty line [200% of poverty]. For Travis County, that’s 373,270 people in 2011.)
From: Selby, Gardner (CMG-Austin) [mailto:email@example.com]
Sent: Monday, February 25, 2013 11:03 AM
To: Eva DeLuna Castro
Subject: RE: use of SAIPE/recent TX PolitiFact
I am still a bit stumped. That is, could it be correct to say, per acs 2011 v acs 2009-11, that due to margins of error, the number of residents in poverty in 2011 was as low as 178,878, compared to as many as 187,789 in 2009-11 (three-year)?
Conversely, could 205,993 residents have been in poverty in 2011 compared to 175,741 in 2009-11?
Maybe I am not adjusting properly for the plus-minus margins.
Big picture: It’s possible poverty was down in 2011, but that’s not statistically certain?
W. Gardner Selby
You’re adjusting correctly. The more years of data that ACS uses for the survey estimates, the smaller the margin of error gets. That’s why smaller geographic areas are only available at the 5-year ACS level. (As you noticed, there’s also 3-yr averages, and 1-yr).
The problem here is, you’re dealing with a range of years after a recession, with much higher-than-average unemployment (and therefore poverty). The further you go back to 2009, the worse you should expect income/unemployment to get. But when surveys require several years’ worth of data in order to be statistically significant, you can’t isolate year to year trends – they all get smoothed out in the averages. That’s why looking at long-term trends, like over a decade, are more helpful. They tell you what’s changing (or not changing) in the underlying nature of an area’s economy that really improves or worsens residents’ quality of life.
Big picture: neither the SAIPE nor the ACS say that anything significantly changed between 2010 and 2011 in Travis County. Longer-term trends show poverty getting worse.