On May 23, the administration released the President’s budget for fiscal year 2018. We now have enough details about the proposed budgets for statistics to start filling in detailed information about cuts.
The Census Bureau is by far the largest of the U.S. statistical agencies. The budget proposes an increase for the Census Bureau of $54 million, or 3.7% for FY 2018. But because of the periodic nature of the decennial population census and the 5-year economic census, the Census budget is hard to understand, and what looks like an increase actually turns out to be a substantial cut relative to what’s needed. What I’m seeing has me concerned.
It’s worth emphasizing that the President’s budget is never the final word on the budget. Appropriations are set by Congress, and they frequently deviate from the President’s budget. My observation, however, has been that Congressional appropriations for statistics are seldom higher than the President’s budget proposals. In the case of the Census Bureau’s budget, I hope that appropriators understand that the proposed budget would severely under-fund the 2020 Census and other core Census Bureau statistics and will decide to boost the appropriation.
2020 Census. The essential background for understanding the budget for the 2020 Census is that mail response rates have declined significantly since 1980, which has driven up the cost of counting each housing unit. The 2010 Census cost $12.3 billion (in 2020 dollars) and the Census Bureau estimated that using the traditional methodologies in 2020 would cost $17.8 billion. To avoid continuing cost increases, the Census created a plan that would limit the total cost of the 2020 Census to $12.5 billion, saving taxpayers about $5.3 billion.
A key part of the plan was upfront investments in address canvassing, optimizing self-response, utilization of administrative and third-party records, and reengineering field operations. The plan required spending money prior to 2020 in order to keep costs down during the census year. The FY 2017 budget, released by the Obama administration in February 2016, included this timeline for the investments that were needed:
The actual 2017 continuing resolution, however, appropriated only $659 million, a shortfall of $122 million. The administration’s 2018 budget proposes an increase to $800 million in 2018, but that not only doesn’t make up for the FY 2017 shortfall, but represents an additional shortfall of $112 million relative to what was planned for a year ago. The cumulative shortfall over these two fiscal years would be about $236 million.
Meanwhile, the GAO has reported that the Census Bureau’s original funding plan may not have been adequate, and the required upfront funding needed to meet the Census Bureau’s goals may actually be higher than originally estimated. One of the risks, of course, is that if the upfront investments fall short, the short-term costs of conducting the Census in 2020 may balloon similarly to past decennial census costs. That would, of course, be bad news for the taxpayer. The Congressional appropriations committees need to take care that the needed investments are fully made so the 2020 Census can be conducted in as cost-effective and accurate a manner as possible.
2017 Economic Census. Unfortunately, my concerns about the proposed Census Bureau budget don’t end with the decennial census. Fiscal year 2018 is also a key year for the 2017 economic census; it’s the time when the actual data collection and follow-up takes place. Keep in mind that the economic census is one of the most critical economic data collections—it serves as a benchmark for not only almost all of the Census Bureau’s annual surveys and monthly indicators, but also is the key input in the benchmarking of BEA’s GDP estimates. In the past, this has usually been associated with a substantial increase in the budget allocated to the economic census. For example, the FY 2013 budget proposed an increase of $41 million to a level of $164 million for the 2012 Economic Census and Census of Governments; the actual appropriation came in at $150 million (an increase of $27 million from the prior year; subsequently, the appropriation was cut further by sequestration). In contrast, the Trump administration’s FY 2018 budget proposes no increase in funding for the 2017 Economic Census and Census of Governments with a proposed level of $110 million. While the budget touts the move to Internet-only data collection as a major cost saving, I still have trouble believing that the quality of the economic census can be maintained with essentially flat funding.
Other cuts. The FY 2018 Census budget also includes flat funding or modest cuts to many other programs, including the American Community Survey, business statistics (such as the the retail trade and services surveys), general economic statistics (such as maintaining the business register and managing the Federal Statistical Research Data Centers), foreign trade statistics, and household statistics (such as the Current Population Survey, which is also partially funded by BLS, and the Survey of Income and Program Participation, or SIPP). Among the specific program changes are redesigns of the Annual Capital Expenditure Survey and the Business R&D and Innovation Survey and a reduction in the sample size for SIPP to about 32,000 households (the most recent panel had about 53,000 households). The budget also mentions cutting back on the follow-up efforts for non-respondents and on the training of field representatives.
High quality national statistics are essential for the smooth operation of both government and business, and Congress needs to ensure that the Census Bureau is adequately funded.
In a future post, I plan to look at proposed cuts to some of the other statistical agencies.