Weathermen and economic forecasters have much in common: They both face public scorn when their predictions of the future go awry. While soggy boots are often a subtle reminder of a weatherman’s incorrect forecast, a major deviation from an economic forecast can change the path of public policy in ways that may amplify economic downturns and fuel unsustainable expansions. So accuracy is of paramount importance.
While many individuals generally perceive weather forecasters as having a poor forecasting record, the reality, however, is that improvements in data collection through radar and satellites, as well as more powerful computers and better forecasting techniques, have noticeably improved weather forecasting over the past 50 years. According to one practical metric, the horizon over which daily forecasts can be accurately made has lengthened from three to more than seven days over that time.
Just as more data can improve forecasts, less data can worsen them. For example, suppose weather forecasters were asked to forecast tomorrow’s temperature without being able to measure today’s temperature. Further suppose that once they actually receive a reading on today’s temperature it would subsequently be revised by, say, plus or minus five degrees. This would surely have a negative impact on weather forecast accuracy – and the same is true for economic forecasts.
Unfortunately, data delays and revisions are a reality for economic forecasters. The most common measure of economic performance, gross domestic product (GDP), is released with a two-month delay, and even after it is released it is subsequently revised by Statistics Canada as it receives new information about the economy.
From a policy-making perspective, having inadequate information on the current state of the economy can delay the implementation of policies that could smooth the economic cycle. For example: In early October, 2008, the federal government, armed with GDP data up to July, 2008 still believed that Canada could avoid a recession, and therefore did not see the need for further stimulus spending.
Because GDP for the current quarter is unavailable, economists need to estimate current GDP before they forecast its future path – a practice that has become known as nowcasting. Traditionally, they would rely on various economic indicators, such as employment or manufacturing data, to nowcast GDP. However, many of these indicators are themselves released with a lag, or are subject to revision, so they still suffer some of the symptoms of GDP itself.
More recently, economic researchers have begun exploiting new electronically recorded indicators that could offer more timely economic guidance. In a new C. D. Howe Institute Commentary, I show that Google searches seem to offer interesting clues about the economy. Growth in the search term “recession” coincided almost exactly with the duration of the 2008-09 recession, while growth in the search term “jobs” follows the path of the observed unemployment rate. Although there remains a lot of work to be done in terms of preparing these data for economic monitoring purposes (such as removing the October, 2011 spike in “jobs” searches due to the death of Apple Inc.’s Steve Jobs), the breadth and timeliness of these data could make them a useful addition to the economic forecaster’s toolbox.
Another data source currently being explored for nowcasting purposes is the use of electronic payments data, such as debit and credit card transactions. Although these record actual spending activity, and therefore are less subject to interpretation errors than Google search data, they are more difficult to follow, because they are not publicly available.
With these new data sources, in 50 years we may look back and say that Internet searches and other electronically recorded data were the “radars and satellites” of economic forecasting. More accurate economic forecasts may help improve public policy decisions that affect the robustness of the economy and strength of domestic job markets.
Greg Tkacz is associate professor and chair of the Department of Economics at St. Francis Xavier University. His report for the C.D. Howe Institute, “Predicting Recessions in Real-Time: Mining Google Trends and Electronic Payments Data for Clues,” can be found at www.cdhowe.org.