Census is counting on ad campaign

It started when a quirky advertisement promoting the U.S. Census, starring actor Ed Begley Jr. and directed by Christopher Guest, aired during the Super Bowl.

Conservatives howled. Sen. John McCain, R-Ariz., said that the $2.5 million ad buy was "symptomatic of the spending practices of the federal government and the Congress in a way that is completely out of touch with what's going on out there in the real world."

Then last week, as advance letters urging Americans to fill out census forms started arriving in mailboxes across the country, some conservatives fumed even more. Why send out a "letter telling me I will soon receive a letter?"

To answer that question we need to consider what the decennial U.S. Census has in common with a survey, and what it doesn't.

Surveys are, by definition, more efficient than a census. We do surveys because it costs less to draw a sample of a population than to count every member. But the Constitution, as interpreted by the Supreme Court, is quite clear: Every 10 years, the government must conduct an "actual enumeration" that counts every person in the country in order to "apportion" members of the House of Representatives.

From 1790 to 1960, the government sent out census takers to knock on doors and ask questions about those living in each household. In 1970, to reduce costs, the Census began mailing out forms to each household for Americans to fill out and mail back. In order to satisfy the mandate to count every person, the Census Bureau then sends out live census takers to follow up at residences that failed to return their forms.

In 1970, the rate of return for paper forms was 78 percent. It dropped to 75 percent in 1980 and 65 percent in 1990, greatly inflating costs. Facing a projected return rate of just 61 percent in 2000, the Census Bureau launched a massive campaign, including paid advertising, to motivate residents to respond. The effort halted the decline and produced a modest increase to 67 percent.

If those were the response rates of a sample survey, we might be satisfied, since telephone response rates on the best political surveys typically fall below 30 percent. But the constitutional mandate leaves the Census Bureau no wiggle room: It must try to count every person, and that drives up the cost.

The numbers are fairly straightforward: All those live census takers are expensive. The Census Bureau estimates that if every household were to return the forms they receive in the mail, it would reduce its marginal costs by $1.5 billion. As such, every 1 percent increase in returns of the paper questionnaire reduces taxpayer costs by $85 million.

So to try to boost that return rate, the Census Bureau is spending $340 million on a promotional campaign, including $140 million for paid advertising, amounts that census communications director Steve Jost tells me are equivalent to what they spent in 2000, adjusting for inflation in paid media.

Many of the complaints about the census Super Bowl ad focused on the ad itself. Entertainment Weekly rated it one of the five worst ads of the night, and it ranked 52nd out of 65 Super Bowl ads in a focus group conducted for USA Today. Time's James Poniewozik wondered if the ad's dry humor would appeal to more than "fans of ironic indie movies." And Republican pollster John Zirinsky, who had tested a potential Super Bowl ad for a client, said the census effort "failed several essential tests" for a successful spot, including credibility and the ability to generate "buzz."

But the goal of the census campaign, according to Jost, was "not to win a competition" on Super Bowl Sunday. The bureau planned its campaign "across television, the Web and social media," and worked "to create a 'water cooler' quality to our initial ads" that would start a conversation about the census in both "viral" and "real terms." And a social media analysis conducted by the Web site Mashable found the census ad ranked in the top five among Super Bowl ads for generating positive chatter on social Web sites like Twitter.

Jost also claims that the overall promotional campaign is hitting its targets. An ongoing daily tracking survey conducted by the census shows that, as of early March, 78 percent of Americans are aware of the census and 84 percent of those who are aware intend to participate, roughly the same levels as recorded at this time 10 years ago.

Whatever one might think about the execution of the Super Bowl ads, their strategy was not created on a whim. To plan the overall promotional campaign, the Census Bureau invested in a massive, high-quality survey (results here and here) over 100 focus groups to test ad themes and concepts. The offbeat approach of the Super Bowl ad resulted in part from evidence of considerable cynicism about government demonstrated by that research.

And why send out advance letters? Because they work. According to Jost, experiments conducted prior to the 2000 Census showed that the letters help boost return rates by 6-12 percentage points.

Does it make sense to spend $14.7 billion (the estimated cost of the 2010 Census) every 10 years trying to get each and every household in the United States to fill out a census form? Probably not. There is a more efficient alternative, but it requires the use of statistical sampling and, presumably, bipartisan support for a constitutional amendment to update the meaning of "enumerate." Too bad the same conservatives who complain about the cost of the census also reject that idea out of hand.