From: Philip Cross
To: Concerned Canadians
Date: October 20, 2017
Re: Policymakers Need Not be Passive Victims of Economic Data Revisions
Policymakers need not be the passive victims of revisions to the economic data they are scrutinizing. They should proactively take measures to anticipate revisions.
To paraphrase Kierkegaard’s observation about life, economic policymaking must be lived forward, but the economy can only be understood backwards, and that backward understanding itself can change due to revisions in data.
Revisions to economic data, with a few exceptions, are part of the normal process of compiling and improving statistics as I discussed in a recent C.D. Howe paper.
Statistical agencies routinely revise data as they are collected, compiled, and disseminated. Preliminary estimates of monthly or quarterly series such as retail sales, trade flows or corporate profits are based on surveys designed to produce a quick estimate of overall trends, with response rates well below 100 percent. As more responses come in over the following few months, estimates are revised. Eventually, high-frequency survey data are supplemented by more comprehensive annual surveys and ultimately are benchmarked to more complete data, usually based on tax records – compensation of employees, for example, is benchmarked to income tax data, which become available with a three-year lag. These more comprehensive data are also authoritative at the industry and provincial levels, where detailed estimates from the preliminary survey are problematic.
However, this does not mean they should be ignored.
The Bank of Canada, for example, found it could predict some revisions to consumer spending and GDP through its monitoring of the use of debit cards. As well, knowing that the GDP estimates might be revised more around turning points means that, during these periods, analysts should build larger confidence intervals around their GDP estimates. Finally, analysts can put more emphasis on data that are less liable to revision, such as employment or retail sales, while downplaying statistics such as exports that are known to be more variable.
Policymakers have expressed their frustration with data revisions, saying that inappropriate policies sometimes were adopted because misleading data were incorporated into the policymaking process. In the United States, the Fed has been particularly vocal in its criticisms; in Canada, the Bank of Canada has been publicly more discreet. GDP is a frequent target of such criticisms, although not even the Labour Force Survey has been immune. As a Bank of Canada analyst has noted, policymakers have to deal with uncertainty about the future, the present, and the past when formulating policy. Revisions mean that the inevitable uncertainty about making projections also needs to take account of the lower, but non-trivial, uncertainty about what happened in the past.
Inevitably, however, analysts must accept that their knowledge of the world will always be uncertain, regardless of the quality of the data. Trying to understand the economy based on data alone risks being so backward-looking that mistakes will be made, irrespective of revisions.
Philip Cross is a Research Fellow at the C.D. Howe Institute and former Chief Economic Analyst at Statistics Canada.
To send a comment or leave feedback, email us at blog@cdhowe.org.
