“The world of statistics has offered me a wonderful career.” Robert Groves looks back on heading the US Census Bureau

Robert Groves is currently the Provost of Georgetown University in Washington, D.C., as well as Gerard Campbell Professor in the University’s Department of Mathematics and Statistics. He is renowned for his expertise in sociology. He previously served as the Director of the United States Census Bureau from 2009 to 2012. Groves obtained an A.B. degree in sociology from Dartmouth College, followed by M.A. degrees in both sociology and statistics and a PhD in sociology, all from the University of Michigan. He then worked as a research professor in survey methodology at the University of Michigan. He was also a researcher in the Joint Program of Survey Methodology, housed at the University of Maryland, College Park.

When Groves was an Associate Director at the U.S. Census Bureau in the early 1990s, he argued that potentially millions of minorities who typically voted Democratic were being undercounted. Groves advocated for the use of statistics to account for this discrepancy. He was later nominated by President Barack Obama in April 2009 to head the Census Bureau. Republican senators raised concerns that Groves would apply statistics to populations believed to be undercounted but Groves discounted this during the confirmation hearings and he took office in July 2009.

Statistics Views talks to Mr Groves about his illustrious career.

thumbnail image:

1. You served as the Director of the United States Census Bureau from 2009 to 2012. What are your memories when you look back on your time in this role and what were your main achievements?

The most important personal benefit of serving in a role like that of director of the Census Bureau is that you meet community leaders and large sets of people throughout the country. I learned much about different racial, religious, and cultural subgroups in the US. I learned about their fears and aspirations for their people.

The staff of the Census Bureau succeeded in mounting a Census for 2010 that performed well within the constraints of the design. Commitment to task and resilience to unexpected events produced a $2 billion savings to the country.

2. How do you think the Bureau evolved during your time there overall and adapted to the changing needs of the statistical community?

We attempted to streamline the management structure of the regional offices; we launched a set of bottom-up processes for new ideas to improve efficiencies in statistical operations; we mounted new work in the use of adaptive designs utilizing real-time data to adjust procedures during data collection.

3. You brought with you an impressive scholarly background to the role of Director after heading the University of Michigan’s prestigious Survey Research Center and you were also a researcher in the Joint Program of Survey Methodology, based at the University of Maryland. After accepting President Obama’s invitation, you said “It was payback time for me.” What did you mean by this and what was it that persuaded you to accept the role?

I have been very fortunate in my life to learn from scholars in my chosen field, to have been surrounded with talented colleagues, and to have taught very bright students. The world of statistics has offered me a wonderful career. At the time I accepted the nomination, there were grave concerns about whether the Census Bureau was capable of completing a successful census. There are several of my friends and colleagues who believed that we could not let this happen. That motivated my acceptance of the role.

At the time I accepted the nomination, there were grave concerns about whether the Census Bureau was capable of completing a successful census. There are several of my friends and colleagues who believed that we could not let this happen. That motivated my acceptance of the role.

4. With an educational background in sociology from Dartmouth College and University of Michigan, when and how did you first become aware of statistics as a discipline?

I became aware of statistics first through application, to social science survey data. This was an undergraduate class taught by James Davis, the father of the General Social Survey at NORC. He threw a data set at us and told us to pose questions to it that interested us. I loved it. The next summer I found myself as a programmer in one of the first interactive statistical software systems, programming a factor analysis. I didn’t know what I was doing, but I fell in love with everything about it. I then found an important mentor in Leslie Kish who taught me much about statistical design principles that are ageless.

5. You are the author of several books on surveys such as Surveying Victims and Survey Errors and Survey Costs. What is the most common error made in surveys from your experience?

Perhaps the most common error is that people assume surveys are simple. One merely has to think of what questions they have in mind, find some people to answer them, and then simply summarise the results. We now know that surveys are assemblies of thousands of details, each of which can affect the cost and error properties of the survey results. Sampling frames are complicated beasts subject to duplication and missing units; selecting and identifying the sample is not intuitive; constructing questionnaires that diverse people can understand; contacting and getting cooperation of the sample units is fraught with difficulties; processing the answers into analyzable forms is laborious and complicated; and then it’s difficult to perform statistical analyses that properly reflect the inferential potential of the survey.

6. You supported the use of sampling to adjust an undercount in 1990 but the Supreme Court ruled that methodology could not be used to apportion House seats. It can be used to draw district lines and to distribute tax dollars, though. It seems there is quite a tight line to walk between the Census and politics but an article from The Washington Post suggested that your scholarly background helped you to win over your critics?

My current belief on statistical adjustment to the Census is that the adjustment tool must be of higher quality than the census being adjusted. A very, very good census is difficult to make better with statistical adjustment. A bad census can profit from adjustment. Further, it’s easier to have useful adjustment when the adjustment tool is quite different from that being adjusted, other things being equal. The weakness of traditional adjustment tools is that the basic data collection method is very similar to that of the census itself.

7. It is not news that survey response rates have been dropping for years. Given that marketers, managers, and pollsters rely upon survey data, how can we increase the response rates for surveys?

After decades of thinking about this problem, I believe that for many surveys the causes of nonresponse are diverse. In some surveys the causes have little correlation with the attributes being measured. Hence, the nonresponse produces little damage to estimates. The problem with nonresponse is that the causes of nonresponse are not themselves measured, and we cannot distinguish when it hurts quality and when it does not.

My current belief on statistical adjustment to the Census is than the adjustment tool must be of higher quality that the census being adjusted. A very, very good census is difficult to make better with statistical adjustment. A bad census can profit from adjustment. Further, it’s easier to have useful adjustment when the adjustment tool is quite different from that being adjusted, other things being equal. The weakness of traditional adjustment tools is that the basic data collection method is very similar to that of the census itself.

8. You are now the Provost of Georgetown University, as well as Gerard Campbell Professor, Department of Mathematics and Statistics. Are you still able to carry out research interests in survey methodology?

Thus far my duties have not permitted time for my research interests.

9. What has been the most exciting development that you have worked on in statistics during your career?

I think the interface between statistical adjustment for nonresponse and causal thinking about nonresponse was the most stimulating. My colleague, Trivellore Raghunathan, and I led a group of graduate students through this literature; created a peer-review paper from it and motivated a set of linked research projects. It was exciting to me.

10. What do you think the most important recent developments in the field have been? What do you think will be the most exciting and productive areas of research in statistics during the next few years?

I think the developments on high-dimensional data, algorithms for analysis coming from computer science, and the prospect for conjoining data from multiple sources have great potential for helping the world with better information. The combination of naturally occurring massive data (often through internet-related sources), survey data, and other administrative data sources has great potential.

11. What do you see as the greatest challenges facing the profession of statistics in the coming years?

Statistics is changing because data are changing. Computationally-intensive approaches to data require new approaches to traditional statistics. The field needs to be open to new approaches.

12. Are there people or events that have been influential in your career?

Leslie Kish, a sampling statistician and broad thinker; Charles Cannell, a psychologist and data collection oriented social scientist; and Robert Kahn, a social psychologist and great mentor to many.

 

Copyright: Image appears courtesy of Mr Robert Groves