The limitations of survey data are reflected in the political adage that, “the only poll that counts is on Election Day.” In other words, real numbers trump survey projections.

Even so, surveys can still add value by offering insights not otherwise inferable from the available hard data. For instance, election and voter registration data do not include demographic information, such as on voter incomes or education. However, a poll that found, say, 30 percent of likely voters have a college degree, could still be correct on that point even if it failed to accurately predict voter turnout or the winner of the race.

The same holds true in other areas as well, such as the health insurance coverage estimates released today by the Census Bureau — which are derived from its Current Population Survey Annual Social and Economic Supplement (CPS ASEC).

The Census data does offer some useful insights about how various demographic factors correlate with different sources of health insurance coverage, or the lack of coverage.

Yet, when it comes to measuring the number of people with each type of coverage, and how coverage patterns change over time, the better source is the actual enrollment data from private plans and government programs.

That caution is particularly relevant to the estimates released today by the Census Bureau. Because those estimates are for 2014, the first year of implementation of the Affordable Care Act (or Obamacare), many will be looking to the Census numbers to determine the effects Obamacare on health insurance coverage.

Comparing the new Census estimates with actual enrollment data shows the degree to which the Census estimates are either on or off target.

The Census report estimates that the number of uninsured people decreased by 8.8 million in 2014. That estimate is close to the figure derived from enrollment data for private plans and the Medicaid program—which show that the number of Americans with health insurance increased in 2014 by 9.2 million individuals.

However, where the estimates reported by Census miss the mark are in the composition of the coverage changes. Specifically, the Census report estimates that Medicaid grew by 6.7 million individuals in 2014, yet enrollment data from the Centers for Medicare and Medicaid Services show that enrollment actually expanded by 9 million individuals.

In contrast, the Census report estimates that private market coverage grew by about 11 million individuals, when in fact net private market enrollment was nearly flat for the year. The discrepancy appears to mainly be the result of Census significantly misestimating the size of the individual (or “direct purchase” in Census terminology) health insurance market.

Bottom line: Health insurance enrollment data show that the total number of Americans with health insurance increased in 2014 by over 9 million individuals. However, the enrollment data also indicate that almost all the growth occurred in Medicaid, as there was relatively little net growth in the number of people with private health insurance coverage. The Census estimates are consistent on the first point, but miss the mark on the second.