We’re trying to count people but we’re not coming up with numbers that make much sense - or differences that are troubling. Let’s back up. We’re trying to count people by educational attainment. The following is for those 25 years and over:
Based on the 2010 American Community Survey:
From the 2010 Digest of Education Statistics (based on the Current Population Survey):
These differences cannot be accounted for by sampling error - the margin of error is smaller than the differences between the two surveys in most cases.
Based on the 2010 American Community Survey:
| Estimate | Margin of Error | |
| Some college, less than 1 year | 12,846,799 | +/-80,342 |
| Some college, 1 or more years, no degree | 30,622,369 | +/-89,558 |
| Associate's degree | 15,553,106 | +/-65,380 |
| Bachelor's degree | 36,244,474 | +/-119,630 |
From the 2010 Digest of Education Statistics (based on the Current Population Survey):
| Estimate | Margin of Error | |
| Some college | 33,662,000 | +/-186,900 |
| Associate's degree | 18,259,000 | +/-142,700 |
| Bachelor's degree | 38,784,000 | +/-198,100 |
These differences cannot be accounted for by sampling error - the margin of error is smaller than the differences between the two surveys in most cases.
Presumably, the larger margin of error in the CPS March supplement data is due to its smaller sample size. Also the March Supplement is a monthly sample (actually collected over 3 months) versus the ACS which is collected over the course of the year. The ACS website states:
The ACS seems to be a more complete sample in terms of its coverage (both in geography and time frame) yet defers to the CPS when totals are being reported (based on above statement). Moreover, the CPS continues to be the official source for poverty estimates. A guidance and fact sheet are available as well as some differences between CPS and ACS.
As far as I can tell Census has not made any attempt to reconcile the differences and I don’t really know if they can. I would assume the ACS numbers to be superior in terms of educational attainment and anywhere where I would think of numbers on an annualized basis. The deference to the CPS in terms of poverty seems from the outside to be a bureaucratic wrangle between two divisions and/or a desire to preserve continuity in time series.
The strength of the ACS is in estimating characteristic distributions. We recommend users compare derived measures such as percents, means, medians, and rates rather than estimates of population totals.
The ACS seems to be a more complete sample in terms of its coverage (both in geography and time frame) yet defers to the CPS when totals are being reported (based on above statement). Moreover, the CPS continues to be the official source for poverty estimates. A guidance and fact sheet are available as well as some differences between CPS and ACS.
As far as I can tell Census has not made any attempt to reconcile the differences and I don’t really know if they can. I would assume the ACS numbers to be superior in terms of educational attainment and anywhere where I would think of numbers on an annualized basis. The deference to the CPS in terms of poverty seems from the outside to be a bureaucratic wrangle between two divisions and/or a desire to preserve continuity in time series.
No comments:
Post a Comment