My Two Census

Formerly the non-partisan watchdog of the 2010 US Census, and currently an opinion blog that covers all things political, media, foreign policy, globalization, and culture…but sometimes returning to its census/demographics roots.

Posts Tagged ‘2000s’

Transcript from Census Bureau Press Conference

Friday, April 30th, 2010

Here’s the transcript from Census Director Robert M. Groves’ press briefing on Wednesday. Unfortunately, my microphone (on Skype) cut out at the moment that I hoped to ask Dr. Groves a series of questions, which concerned why there are still failures of the paper-based operations control system and who (if anyone) is being held accountable for these errors. I forwarded these questions to Stephen Buckner at the Census Bureau’s Public Information Office, but have still not received a reply. Dr. Groves did briefly acknowledge errors in the PBOCS during his speech, but it is unlikely that any members of the mainstream media who don’t cover 2010 Census operations regularly would know about such a system and what those errors mean. Dr. Groves said, “We continue to struggle with the software system called the paper base operation control system, but we passed, just amazingly, a wonderful threshold last week where we printed out assignments for all these enumerators. It worked. We have assignments ready for 600,000 people who are ready to hit the streets on Saturday. So we’re proceeding. Not that it is the most loved piece of software in the Census Bureau, but it’s working well enough to get the census down so far.”

STEPHEN BUCKNER: Good afternoon. Welcome to the Census Bureau’s news conference on the mail participation rates. I’d like to welcome everyone here in the room, and also those joining us online and via telephone. If you take a few moments, we have some information in your press kits, and also available online, including all the charts that Dr. Groves will be going over today as he walks you through America’s accomplishment in the mail participation rate for the 2010 census as we start to go door to door later this weekend.

Following Dr. Groves’ remarks, we will have a brief Q&A session for the media. Please state your name and organization prior to your question. We’ll try to get to as many questions as we can during the news conference. And with that, I’ll turn it over to Dr. Groves. Thank you, Dr. Groves. (Applause)

DR. ROBERT GROVES: This is the first time I’ve heard applause. Well, welcome. I’m happy to be here and happy to see friendly faces in front of me. Today is a big day for us because we announce the end of the first half of the 2010 census. And we have good news, because we can thank the American public for really the first major achievement of the 2010 census, I think. So first, I need to say why are we honoring the American public in this way, and why is this a notable achievement? You need to know something about levels of participation in surveys in this country, and in fact in the western world over the past few decades to understand how wonderful what happened really is.

For the last 20 years, response rates, the level of participation of the public in sample surveys, in all sectors, the commercial sector, the government sector and the academic sector, have been falling. In fact, in the Census Bureau’s own survey, take the American Community Survey, this very large thing that we do continuously, we’ve lost 5 percentage points from the 2000 experience to now. So when I took this job, I really expected that any achievement close to the 2000 rate was beyond each. In fact, I urge you, urge the journalists here, to call up your favorite survey researcher and ask them one question: could you achieve the response rate today that you received ten years ago on the same survey? And see what they say.

So we had low expectations on getting close to where we were in 2000. And where were we on 2000? The combined short form and long form participation rate in 2000 was 69 percent at the time we cut off for the non-response follow-up. So if we believe those lower rates, we would have expected something lower than that. The short form only rate, out of the 2000, the portion of the households that got the short form, their participation rate was about 72 percent. We chose that as a stretch goal. We were preparing for response rates between 65 and 72 percent in our simulations. Well, what happened was the American public hit that stretch goal, and it was a wonderful display, we think, of civic participation. And I can tell you, the folks at the Census Bureau are dancing down the hallways.

There are a lot of neat things about that; 28 states met or exceeded their 2000 rate, that’s cool. Some of these are pretty large states; Florida, Georgia, Illinois, Michigan, New York, Pennsylvania and Texas. In addition, the District of Columbia and Puerto Rico exceeded their 2000 rate. There are 11 other states within 1 percentage point of their 200 rate. North Carolina and South Carolina are kind of the poster children of 2010. They exceeded their rate in 2000 by 8 percentage points, just off the charts, so as they say these days, need a special shout out, I think. And there are thousands of jurisdictions around the country that have exceeded their 2000 rate.

Let me just give you a sense of this. In your press kit, I think you have notes that seven of the ten most populous counties equaled or surpassed their 2000 rates. Eight of the ten most populous cities did that, likewise. So once again, this is, I think, a congratulations due to the American public for their act of civic participation. And we are grateful and happy to report that.

It might be good to ask the question of why, why was that rate achieved? And the first thing that I want to point out are some hard data we have. The first unambiguous result in the why, I think, has to do with the replacement form. So if you look at this chart, let me help you read this chart. And for those of you on the phone, this is a chart that’s labeled 2010 average mail back participation rates for census tracks by assignment of 2010 replacement form. So it’s a busy graph, I’ll help you read it.

The X axis here is a day, sort of the life cycle of the mail out/mail back process. The Y axis is the participation rate. So the first thing you’ll see is that all of the lines rise as you go from left to right as the mail back rate built over days. Now, there are a lot of different lines here. The solid lines are from the 2010 data. The dashed, or dotted, lines are from 2000. There are three groups of lines color coded. The red line are those tracks that got replacement forms in a blanket fashion. If you lived in that track, whether you returned the form or not, you got a replacement form. The green line are tracks that were targeted. If you yourself did not return a form, you got a replacement form. If your neighbor did, they did not. And then finally, the blue line are the tracks where no replacement form was given.

Look at the dashed lines first. We targeted for blanket treatment the toughest tracks. They had low response rates in 2000. That’s the lowest line, the red dashed line, is low. And the green line were the targeted tracks. They had sort of medium participation rates. But you can see, both of those are pretty far below the blue. So we targeted the replacement form to those areas that tended in 2000 to have low participation rates.

What happened? Well, this replacement form worked. It’s just unambiguous in the data. Let your eyes focus on the solid red line. Look how after the blanket replacement mailing was complete around the 3rd of April, it starts pumping up. It just gets higher than the dashed line and stays higher. And then notice how the green line, solid green line, starts getting higher after the target replacement form. In fact, I can tell you, the green line is still pumping out cases as w speak. This is a good thing for the 2010 census because now let your eye go to the dashed lines and just look at the difference in height between the blue dashed line and the red dashed line. That variability in participation rate is not good for a census. We’d like everybody to have the same participation rate. We squeezed together those lines to this replacement form.

So the first question, why did we get a good participation rate, is the replacement form worked, and it worked in a wonderful way to reduce the variability in response rates. There’s another reason, and we have some data on this but it’s not as hard as these data; we think the advertising and the partnership worked. I think in prior discussions we’ve had, we showed you how the awareness of the census rose dramatically as the advertising and partnership campaigns evolved. That seems unambiguous, that that was a good reason why.

And then there are other things that have to do with judgments. I just told you that the participation rate in 2000 for the combined short and long form was 69 percent. And for the short form only was 72 percent. We’re pretty sure the short form was a great idea based on that comparison. We don’t have an empirical comparison of the long and short form in this census, but one clear, I’d be willing to make this speculation, one clear answer to the question why a high rate of participation rate, or why a high rate of participation, is that the short form reduced the burden on the American public and they cooperated at higher rates.

We are going to do tons of other analysis over the coming weeks to figure out other answers to the question why did this seem to work. That’s our job in order to prepare for the 2010 census. But those are the early findings on the reasons for success.

I want to turn to another issue, and that is the patterns of response rate. You can see on the map that we showed earlier that there’s variability in the response. If you just look at the colors, the colors are related to different response rates. Red is really good, blue is bad, and you can see how the colors vary. We’ve been publishing this map every day on the web, a lot of you have been watching it. You start looking at a map like that and you begin to make up hypotheses about geography as the cause of the participation rate. We’re pretty sure that’s a misinterpretation of this. And I want to give you a sense of the patterns of response.

You know that for several decades, the Census Bureau has tracked differential under counts. When the whole census is finished, differential under counts by race, ethnicity and age, and the patterns are very clear over decade after decade, that younger people, that minority groups, tend to be disproportionately missed in the census. I can tell you day by day, we were watching those differences on participation rates in this first half of the census to see if we were seeing those patterns.

The classic patterns emerge in 2010, they’re there. But the biggest drivers in the participation rate are not race and ethnicity in the 2010 census, but a variety of other indicators. And I wanted to give you a sense of what those look like. A lot of them are indicators of the socioeconomic status of the area. So let me describe what this is. This breaks the census tracks. There are about 65,000 census tracks in the country, into four groups of equal size by the rate of vacancy in the area. What proportion of the houses were vacant based on American community survey data over the past three years? So the X axis here separates all of the census tracks into four groups, from low vacancy rate to high vacancy rate.

The Y axis here is the participation rate. The pattern is beautifully monatomic, as they say, in the statistics business. It goes down with each added quartile from 76 percent way down to 64 percent. Tracks that have high vacancy rates tended to have low participation rates. And it’s a pretty strong effect.

Let me show you the next one. This does a similar sort of thing. It breaks the tracks into four groups by their level of percentage of multi-unit structures. And on the X axis, you see what those rates are. Similar sort of finding; pretty strong effects from 77 percent participation rate down to 64 percent. Tracks that have a lot of multi-unit structures or mobile homes versus those that have single family structures had lower participation rates.

Let’s look at the next one. Renter occupied housing units, same sort of divisions on the X axis, participation rate on the Y axis, moving from 77 to 64. Census tracks, neighborhoods with a lot of rental units had lower participation rates. Let’s look at the next one. Education. We can identify the proportion of people on the tracks with less than a high school degree, same sort to pattern. Tracks that have a lot of people with low education tend to have lower participation rates. And is that the final one? One more, poverty. The American community survey allows us to estimate the proportion of households under poverty threshold; 77 to 64, same thing.

Now in your press kit are other graphs that look at ethnicity and race. You’ll look at those and see smaller effects across those groups than you see across the socioeconomic indicators. Was this present in prior censuses? I suspect it was. Did we have the right data to do this sort of analysis? Now, this is a wonderful benefit of the American community survey that we can track in almost real time other indicators. These are the strongest drivers to participation rate, not the race, ethnicity indicators. But they’re pretty powerful drivers, as you see.

So we can say that tracks that have high rental units, low education, low income, are disproportionately where our non-response follow-up workload is. Those are the neighborhoods that we will be visiting disproportionately in the coming weeks.

I want to say just a bit about our current operations and then open it up for questions. We are proceeding on all sorts of operations now. I’m happy to report that every operation we’ve done since addressing canvassing in summer of 2009 are on time and on budget. Those are going well. We continue to struggle with the software system called the paper base operation control system, but we passed, just amazingly, a wonderful threshold last week where we printed out assignments for all these enumerators. It worked. We have assignments ready for 600,000 people who are ready to hit the streets on Saturday. So we’re proceeding. Not that it is the most loved piece of software in the Census Bureau, but it’s working well enough to get the census down so far.

I want to remind everyone that we have another press conference on Monday. And this is really a press conference to look forward to the non-response follow-up process. It’s a gigantic effort. It’s a complicated effort. We want to make sure you have all the facts that we do about how it’s going. We’ll talk about that. It’s important to note that although we report this mail back rate as if we’ve reached a final stage, this is really the end of the first half and that the census is not over. We will have a disposition on 100 percent of the units that we have on our list, those who didn’t get forms will be visited. Those who got forms and didn’t mail them out will be visited. We will collect information on everyone before we’re through, and that’s the second half that we’re now beginning. So I want to stop at this point and entertain questions. (more…)