What are the problems in producing summary productivity statistics?

View the most recent version.

a) With the concepts?

The Canadian Productivity Accounts produce several different measures of productivity growth. While partial productivity measures are relatively simple to understand, they have been surpassed in the world of practical analysis by the more complex multifactor productivity (MFP) measure. While this has now become the standard among experts,⁹ the MFP is an analytic construct and is derived using specific assumptions regarding the nature of the economy. Being an analytic construct, the concept of multifactor productivity is more difficult for the less expert users of Statistics Canada's products to understand. And the assumptions embedded in the growth accounting framework mean that its validity in the eyes of some users relies on the acceptance of these assumptions.

The Canadian Productivity Accounts have responded to these issues by providing detailed descriptions of the methodology used in developing the measures,¹⁰ and examining the extent to which alternate approaches yield significantly different measures of MFP growth.¹¹

The second major problem with multifactor productivity estimates is that they capture what we cannot explain: they are a residual calculated after other measurable factors have been taken into account. To some analysts, this is not a problem since they want a measure of the externalities that are bestowed on an economy by disembodied technological progress. But even here, guidance is needed on what the underlying factors might be that are behind this component— changes in plant scale or production run length economies, firm reorganizations relating to offshoring and outsourcing, new technologies, intangible capital. To meet demands in this area, Statistics Canada has responded with studies using business microdata in each of these areas.¹²

b) With measurement?

In an economy as large and diverse as that of Canada, it is a Herculean task to calculate a summary statistic for productivity that, in 2005, sums up the efforts of 16.2 million workers, employed in thousands of establishments that produce about $1.4 trillion in output. Statistics Canada does so in its productivity program, which uses an integrated set of data sources produced by the System of National Accounts.

Statistics Canada produces productivity statistics as part of a regular production program. It is not something done as in many other countries, as an occasional research exercise. The production process for the Canadian Productivity Accounts is embedded within the System of National Accounts. The Canadian Productivity Accounts play an important role as an integrator of data from different sources within the agency.

Statistics Canada's integrated national accounts provide the foundations on which the productivity accounts are based. Because they are integrated across several dimensions—from the demand side, from the income side and from the industry accounts—along with detailed input–output tables, there is a solid foundation on which the productivity accounts builds. For example, estimates of productivity using the demand side are reconcilable to those coming from the industry side.

The Canadian Productivity Accounts puts together an integrated set of data on outputs, inputs, labour and capital contributions to the production process. Statistics Canada's Productivity Accounts build first off an integrated set of production accounts—that generate gross domestic product from final demand, and at the industry level with one set of integrated, coherent accounts. The Productivity Group takes this integrated set of accounts and produces a set of estimates of labour services and capital services that are coherent with the output estimates. For example, on the labour side, the Productivity Group chooses amongst various source data (there are multiples sources, i.e. household versus employer surveys, each giving different estimates of labour inputs), ensures the boundaries of the labour sources agree with the boundaries of the industry data, and produces a set of labour inputs (by estimating jobs and hours-worked separately and then multiplying them together). In the case of capital services, the Group takes investment data from a survey of investment, reconciles and modifies them to accord with System of National Accounts boundaries, and then estimates capital services making use of rates of return that are derived from the System of National Accounts estimates of profits or surplus taken from the input–output tables.

Statistics Canada's productivity program also provides quality assurance across all input sources by improving the overall coherence of these products. Analysis in the productivity program, as is the case elsewhere in the National Accounts, is an extension of the particular nature of the production process. The production process in the Canadian Productivity Accounts combines data from different sources. To construct official data series, this production process confronts data from one source (for example industry value added) with data from another (for example, labour inputs). In the end, this comparative process serves to bring a variety of sources into coherence with one another. Data that are generated from production surveys are subject to both response and non-response errors. By examining how one series compares to another (for example, how employment estimates from the Labour Force Survey compare with those from the Survey of Employment, Payroll and Hours), analysts in the Productivity Accounts can assess whether the survey error in one or the other data source is particularly large in one period and adapt the estimate that is most appropriate for the creation of a time series that is not only consistent over time but is also coherent with the other data that are being used in the estimates of productivity.

The Productivity Accounts develops and maintains a large database in support of the productivity program—what some refer to as the KLEMS (Capital, Labour, Energy, Materials and Services) database. KLEMS integrates time series data on gross output, materials inputs, service inputs, energy purchases, labour, investment and capital. Each of these data series is calculated in both nominal dollars and real (constant) dollars. Price indices are collected for each of these series. Finally, KLEMS classifies these series using four different levels of aggregation—corresponding to the S, M and L levels used in Statistics Canada's Input–Output Accounts.

 

9. See Economist 2004.

10. See Productivity Growth in Canada, Catalogue no. 15-204, and the various publications in the series Canadian Productivity Review, Catalogue no. 15-206.

11. Baldwin, Gaudreault and Harchaoui 2001 examine the use of parametric as opposed to non-parametric techniques. Baldwin and Gu 2007a examine the impact of using alternate techniques of estimating the cost of capital when deriving estimates of capital services. Statistics Canada 2007c examines the impact of alternate depreciation rates.

12. These papers can be found in the Economic Analysis (EA) Research Paper Series, the Analytical Studies Branch Research Paper Series and The Canadian Productivity Review, and are summarized in the Update on Economic Analysis section on Statistics Canada's website.