b'InsightsGiven your role and responsibility as chief statistician of theretail sales, you can start to look at data from companies that U.S., what would you say our top challenges are and howaggregate credit card records, because more and more have you sought to address those challenges?purchases are on credit cards. The Census Bureau and the Bureau of Economic Analysis have done research in using A top challenge is the pace of change aroundaggregated credit card records to calculated retail sales. The information and the ways in which information can be madeindividual purchases are de-identified because theyre available. Things have changed considerably in the lastaggregated, but you can see what was purchased using credit 10-15 years. My challenge is to make sure that the federalcards in Chicago or in New York the day after the purchases statistical system stays relevant in an ever-changingtook place. Thats how fast the data are aggregated. You no environment. A common perception of a statistician islonger have to go to the businesses to ask about sales, someone with a green eyeshade who is calculating variancebecause you can see the sales from the purchase end. But you and standard deviations. This is not an accurate portrayal;need to be careful that you are not missing sales that are paid statistical activity really defines information that is used tofor by means other than credit cards in the released indicator. describe groups, even though it comes from individuals. Its business data or social data.What big initiatives are you working on?If you want high-quality information, its very important thatData sharing between agencies, and safeguarding that you think about the mature system of quality measurementdata, is one. There are large amounts of information that the that the statistical community has developed over decades.government has already collected on people. It resides not In some quarters, I think there is a view that if you havejust in Social Security records, but also Medicare and enough data, youre going to get to the right answerMedicaid, veterans records, housing records. Why would you eventually. But a lot of people use big data sets that arentspend all this money to go out in a survey and recollect the complete, that have biases built into them. The statisticalinformation if the information that you wanted has already system has a very mature framework that is important to use.been collected? Doing more data sharing between agencies is How do we take these traditional statistical methods that relya big focus and a key requirement of the Foundations for primarily on surveys and modernize them for using otherEvidence-Based Policymaking Act (Evidence Act). Another types of data? Its a big challenge.priority is to safeguard it at the other end and make sure that youre protecting confidentiality and privacy. That also As the world produces better, faster, more granularincludes assuring that these sensitive data are only used for statistical products, what do you worry about?statistical purposes and not to identify or take action against any individuals. These are two big strategic initiatives that Re-identification. Protecting confidentiality is a big, bigwere working on cross-agency, along with the federal data challenge these days because technology, computing power,strategy. and the availability of open data really create a different environment than we had 30 years ago. The intake side is aWould you tell us more about the federal data strategy? big challenge in terms of new data sources and the rapidity at which you can create products to meet increasingTo help agencies leverage their data as a strategic asset, demands for timely and granular data.the federal data strategy includes four components. These components are the building blocks and guides for federal What are your strategic priorities? How have external trendsagency actions over the next several years. informed and shaped your strategic direction? The first component is enterprise data governance. It includes One of my key priorities is to modernize the datastandardizing metadata, creating inventories, safeguarding collection methods in order to be able to get data out faster.confidentiality and privacy, and so on. The more expansive Surveys take a long time to process and theyre expensive.governance vision includes collaboration across agencies Also, people increasingly dont like to answer surveys. Its anand agency program silos in order to bring multidisciplinary intrusion. Its hard to collect information that way. We alsoexpertise together to formulate and address the big questions have a proliferation of data accessible in less traditional waysthat have been so difficult for agencies to tackle. To be that can be used for statistical purposes. For example, if yousuccessful means changing federal agency cultures not only to are releasing a monthly retail sales economic indicator andask priority questions that are meaningful and specific to the you want to put it out faster than, say, six weeks after youagency, including operational and mission-strategic questions, complete each monthly survey asking businesses about theirbut also to share data across silos within and across agencies. WINTER 2019 / 2020 IBM Center for The Business of Government 41'