-
TEXT
This article reviews recent work that uses principal-component analysis to extract information common to indicators from the Bank of Canada’s Business Outlook Survey (BOS). The authors use correlation analysis and an out-of-sample forecasting exercise to assess and compare the information content of the principal component with that of responses to key individual survey questions on growth in real gross domestic product and in real business investment. Results suggest that summarizing the common movements among BOS indicators may provide useful information for forecasting near-term growth in business investment. For growth in real gross domestic product, however, the survey’s balance of opinion on future sales growth appears to be more informative.
Since the autumn of 1997, the Bank of Canada's regional offices (located in Halifax, Montréal, Toronto, Calgary, and Vancouver) have conducted consultations with businesses across Canada on a quarterly basis. These consultations are now referred to as the Business Outlook Survey (BOS). The BOS provides a timely source of information on what businesses are experiencing and planning. Business consultations are timed to feed into the decision-making process that precedes the Bank's fixed dates for announcing monetary policy decisions. The consultations are structured around a questionnaire. Every quarter, 100 firms that reflect the diverse composition of the Canadian economy in terms of region, type of business activity, and firm size are interviewed. Because the BOS is a relatively new tool, the survey time series is short. The assessment presented in this paper, which is based on charts and correlations, is intended as an initial guide. The findings suggest that the survey serves as a barometer of the Canadian economy and provides leading signals of future activity. The interview responses also inform the Bank about production-capacity constraints, labour shortages, and inflation expectations.
Since the autumn of 1997, the regional offices of the Bank of Canada have conducted quarterly consultations with businesses across Canada. These business consultations are timed to feed into the decision-making process that precedes the Bank‘s fixed dates for announcing monetary policy decisions.
Starting with the winter 2007–08 survey, the results of two additional questions became included in the Business Outlook Survey (BOS) publication: the balance of opinion on past sales and the balance of opinion on credit conditions. This backgrounder briefly describes the two questions and presents the correlations between the responses and relevant economic data.
A number of central banks publish their own business conditions survey based on non-random sampling methods. The results of these surveys influence monetary policy decisions and thus affect expectations in financial markets. To date, however, no one has computed the statistical accuracy of these surveys because their respective non-random sampling method renders this assessment non-trivial. This paper describes a methodology for modeling complex non-random sampling behaviour, and computing relevant measures of statistical confidence, based on a given survey's historical sample selection practice. We apply this framework to the Bank of Canada's Business Outlook Survey by describing the sampling method in terms of historical practices and Bayesian probabilities. This allows us to replicate the firm selection process using Monte Carlo simulations on a comprehensive micro-dataset of Canadian firms. We find, under certain assumptions, no evidence that the Bank's firm selection process results in biased estimates and/or wider confidence intervals.
While a number of central banks publish their own business conditions indicators that rely on non-random sampling, knowledge about their statistical accuracy has been limited. Recently, de Munnik, Dupuis, and Illing (2009) made some progress in this area for the Bank of Canada's Business Outlook Survey (BOS) by estimating the impact of the Bank's non-random sampling on the accuracy of the survey results. They found no evidence that the Bank's firm-selection process results in significantly biased estimates and/or wider confidence intervals than in the random-selection case. The author deepens and extends this work by (i) outlining the statistical properties of population-proportion and balance-of-opinion questions, and demonstrating how their design affects the calculation of the confidence intervals; (ii) examining the variation in statistical confidence associated with changes in the underlying response distribution using actual quarterly BOS results; (iii) considering the possibility that statistical accuracy varies across questions; and (iv) investigating whether the statistical accuracy of the survey results changes with variations in the business cycle. The main findings are that confidence intervals around the population-proportion questions are about half of those for the balance-of-opinion questions, and that the confidence bands around both types of question can change from survey to survey when the underlying response distribution becomes more or less concentrated in particular response categories (such as “higher,” “the same,” or “lower”). The author finds that confidence intervals around the BOS population-proportion questions become somewhat narrower during periods of recession, while those for the balance-of-opinion questions vary within a similar range across the cycle.
About one-half of Canadian firms surveyed by the Bank of Canada’s regional offices between September 2003 and December 2004 reported being adversely affected by the appreciation of the Canadian dollar. Roughly one-quarter reported a favourable impact, and the remainder said that there was no effect.