Government Information in Canada/Information gouvernementale au
Canada, Volume 3, number/numéro 1 (summer/été
1996)
Which Database and Which Estimates for Forecasting Economic Performance? 1
Economist
This paper shows that there is really no way of reconciling
differences in the estimates of Saskatchewan's GDP by the federal and
provincial statistical agencies short of doing them again,
independently. Even then it is not clear how much difference that would
make. There is also no way, therefore, of reconciling what can be
markedly different forecasts of economic performance. Nevertheless, with
other economic statistics at hand, it becomes possible to interpret the
forecasts more effectively.
Les prévisions portant sur les résultats économiques
dépendent
des estimations du Produit intérieur brut (PIB), mais
différents organismes qui
oeuvrent dans le domaine statistique utilisent différentes
méthodes pour mesurer
les éléments qui interviennent dans l'élaboration de
ces estimations. Cet
article fait appel aux prévisions régionales que la Banque
de Montréal
publie tous les six mois pour le Canada et la Saskatchewan, afin
d'illustrer la façon dont
l'économie de la province peut croître à un taux
égal à deux
fois celui du Canada, ou encore à la moitié ou aux
deux-tiers du taux national,
selon les données estimatives de la production réelle que
l'on utilise--celles
préparées par le Bureau de la statistique de la
Saskatchewan, ou celles de
Statistique Canada. Dans ce dernier cas, on utilise les estimations
fondées sur le
coût des facteurs, ou sur le prix du marché.
Le présent document indique qu'il est en réalité
impossible de
réconcilier les divergences entre les estimations du PIB de la
Saskatchewan établies
par les organismes de statistique provincial et fédéral,
à moins de les recalculer
de façon indépendante; en outre, on ne sait pas quel
écart une telle approche révélerait. Par
conséquent, il est impossible de réconcilier des
prévisions sensiblement différentes en matière de
résultats
économiques. Néanmoins, grâce à d'autres
statistiques de nature
économique, il devient possible d'interpréter les
prévisions de façon
plus efficace.
Forecasts of growth in economic activity depend upon estimates of
Gross Domestic Product (GDP), the mainstay of national income accounting,
but each statistical agency can have its own preferences for measuring the
components needed to construct the estimates, and GDP can be measured in
more ways than one. The Bank of Montreal's mid-year regional outlook for
the Canadian economy raises questions about which of two statistical
agency databases (the federal or provincial government) best covers the
information needed to make the estimates of real GDP in Saskatchewan, and
about which estimates (factor cost or market price) best represent the
changes in economic activity that can be expected. Using the Bank's
forecasts, this paper shows how different estimates of real GDP can give
quite different predictions of economic performance in a Canadian
province. The problem is defined in the first part of the paper. The
alternate estimates are explained in the second. The implications of using
one or other of the different estimates to make forecasts are explained in
the third. Introduction The rate of growth of
real or constant dollar, Gross Domestic Product (GDP), is often used to
describe the rate of growth of economic activity in general. Most
economists supplement it with figures on employment and a variety of other
indicators. Real GDP attracts much attention each year, however, when
forecasters release their projections of its rate and explain what they
think might be expected in the months ahead. The Economics Department of
the Bank of Montreal, for example, in July forecasted that the Canadian
economy (that is, real GDP), would grow by 1.80 percent in 1996 and 3.50
percent in 1997 following an actual rate of 2.20 percent in 1995, while
the Saskatchewan economy would grow by 1.75 and 2.75 percent in 1996 and
1997 respectively, following an actual rate of 0.80 percent in 1995 (Bank
of Montreal, July, 1996). The Bank's projections for Saskatchewan
raise three questions: Which estimate of real output should be used to
make the provincial forecasts? Is the rate of growth of the provincial
economy likely to be above or below the national rate in 1996 and 1997? Is
a forecasted rate of growth of real output a good proxy for what might be
expected from economic activity in general? There are three reasons
for being concerned about the Bank's forecasts. First, they are for real
GDP at market prices in the case of Canada, but for real GDP at factor
costs in the case of the provinces. Secondly, in the case of Saskatchewan,
there are two sets of estimates of real GDP at market prices which imply
different things about this year and next, one made by Statistics Canada
and one by the Saskatchewan Bureau of Statistics. Thirdly, forecasting
models usually implicitly assume that most of output growth is transformed
into growth in incomes, spending, and employment in the same time period
and in the same economy. That may not be what happens in an export,
primary product economy like the one in Saskatchewan.
Estimates of Real Output The Bank of Montreal's
forecasts refer to real GDP, but in two forms: real GDP at factor cost and
real GDP at market prices. As long as there is a constant difference over
time between a market price and a factor cost estimate of real GDP, there
will be little difference in the rates of growth derived from them, but
there will be different annual rates when the difference changes over
time. Market price GDP is the total value of output from the point of
view of the residents of an economy who are buying the output in the
market. It includes indirect taxes, excludes subsidies, and includes
capital consumption allowances (as well as a statistical adjustment,
usually small, to account for errors in the estimates) because indirect
taxes and capital consumption allowances raise, and subsidies lower, the
price a consumer pays for a product, compared with what he or she would
pay if only the costs of hiring the factors of production (labour,
capital, and land) had to be paid to get the output produced. Factor cost
GDP is the total value of output from the point of view of what has to be
paid to hire all the factors of production needed to produce the output,
that is, taking into account the total factor cost of production. A market
price estimate is usually higher than a factor cost estimate of real GDP.
Factor cost GDP is derived by adding up the value added in each
industry in an economy during a particular period of time such as a year. The
market price estimate is derived by adding up all the incomes that are
received by residents from their participation in production during the
year and, to cross-check, by adding up all that is spent on that
production by residents during the year. The total of all incomes received
from participating in production is equal to all that is spent on the
output. Because forecasting models typically project output from past
spending patterns, and price indexes are developed for spending categories
of goods, they usually project the expenditures side of real GDP at market
prices, as the Bank of Montreal has done for Canada in their mid-year
regional outlook.
Statistics Canada has estimated that the value of real GDP at market
prices in Canada, in every year from 1984 to 1996, was 12 percent above
its factor cost value. There is, therefore, little difference between the
rate of growth in real output that is derived from Statistics Canada’s
market price GDP and the rate derived from its factor cost GDP. It was
just 0.05 percentage points on average each year (differences as high as
0.51 and as low as 0.02). Over the same period, Statistics Canada’s market
price estimate of Saskatchewan’s real GDP was an average of seven percent
above its estimate of the factor cost real GDP for each year. The
difference was as high as $2.2 billion and as low as $422 million out of a
total of $15 to $20 billion. The rate of growth differed by an average of
0.33 percentage points each year, but by as much as 3.09 and as little as
0.15 percentage points in any one year.
To complicate things, in the case of Saskatchewan, there are two market
price estimates of real GDP. The Saskatchewan Bureau of Statistics
prepares its own from essentially the same information that Statistics
Canada uses for its factor cost and market price estimates. The Bureau’s
market price estimate, like the Statistics Canada market price estimate,
was an average of seven percent above the Statistics Canada factor cost
estimate each year between 1984 and 1996, but the Bureau’s annual rate of
growth was an average of 1.14 percentage points higher than the rate of
growth derived from Statistics Canada’s factor cost estimates, as much as
3.90 and as little as 0.21 percentage points higher.
The annual rates of growth of Saskatchewan’s real GDP between 1984 and
1996 are shown in Table 1
with the Statistics Canada national market price rate for comparison.
Statistics Canada’s factor cost estimates have the Saskatchewan economy
growing by an average of 1.69 percent each year between 1984 and 1996
(0.77 percentage points below the average annual national rate).
Statistics Canada’s market price estimates hav eit growing by an average of
2.03 percent each year (0.34 percentage points higher than the factor
cost rate, but 0.43 percentage points below the national rate). The
Saskatchewan Bureau of Statistics’ market price estimates have it growing
at an average of 2.84 percent each year (1.15 percentage points higher
than the Statistics Canada factor cost estimates and 0.38 percentage
points above the national rate). Two of the average rates are lower and
one is higher than the average rate of growth of Canadian real GDP at
market prices, but there are substantial differences in particular years,
as much as 3.09, 3.90 and 5.40 percentage points in specific years.
While the difference between the factor cost and market price values of
total output is composed of indirect taxes, subsidies, capital consumption
allowances, and a statistical discrepancy, the difference between the
Statistics Canada and Saskatchewan Bureau of Statistics market price
values of total output seems to depend upon when the information used to
make the estimates becomes available.
The Saskatchewan Bureau of Statistics introduces three pieces of
information on output into its database more quickly than does Statistics
Canada: the output of agriculture; the output of non-renewable natural
resource products; and the output of government services. Together, they
represented 28 percent of the total output of the Saskatchewan economy on
average each year between 1984 and 1996. The Departments of Agriculture
and Energy and Mines in Saskatchewan keep detailed monthly data on output
in the sectors which fall under their responsibility for policy. In the
case of government services, where provincial classification schemes
differ, the federal statistical agency rearranges the provincial data so
that the reports for each province can be compared, but that means that
it takes longer to integrate the information which comes from the province
into the federal database.
Some differences between federal and provincial estimates of real GDP at
market prices may be a matter of timing, as, for example, when the
accounting basis for a series is changed as it was in the case of
agriculture recently. Other differences can come from the fact that it is
easier for federal statistical agencies to marshal the research resources
needed to develop sets of data such as sophisticated price indexes when there
are economies of scale to be gained by doing research for more than one
jurisdiction at a time.
It is tempting to believe that the Saskatchewan Bureau of Statistics’
estimates of real GDP come closer to describing what is happening to
economic activity in general in Saskatchewan than do the Statistics Canada
estimates. However, the difference between them and the Statistics Canada
estimates should disappear over time as the information from the province
is brought into the Statistics Canada database. Between 1984 and 1996, the
Saskatchewan Bureau’s market price estimate of real GDP was the same a s he
Statistics Canada market price estimate in only one year, 1987, and the
differences were substantial in some years, although they averaged out
over the 12 year interval. It cannot be assumed that they will give
similar forecasted rates of growth for any one year.
Forecasting Economic Activity from Output
Table 2 shows what the
Bank of Montreal’s forecasts for the Saskatchewan economy would have been
if the other estimates of real output had been used in its model to make
the projections. The Bank’s forecased Canadian rate is included for comparison.
Assuming that the Bank of Montreal’s forecasting model accurately defines
the time path of changes in the provincial economy, Saskatchewan could be
growing at more than twice, or 0.64 percentage points below the national
rate in 1996, and at almost twice, or 1.68 percentage points below it in
1997. These rates are for the growth in output which is not necessarily
transformed into spending when the product is ready for the market. With
grain farmers in Saskatchewan currently harvesting the best crops they
have seen since 1986, and good prospects for sales from mining, quarrying,
and oil wells, the 1996 rate of growth of output could be substantially
above the Canadian rate. Whether that is above or below the Canadian rate
in 1997 depends upon the weather and international markets for grains,
potash, uranium, and oil.
The rate of growth of real output in any given year can be a reasonable
proxy for changes in economic activity in general only in so far as the
incomes generated from producing that output are spent in the same year
and on goods produced in the same economy. That need not happen with
primary products. Canadian grain farmers, for example, use inventories of
an unperishable primary product to change their output in any one year
into a stream of income and spending over several, one that better fits
their individual needs and better reflects developments in the
international markets for their product. Grain produced in 1996 will be
transformed into some income in 1996, and into spending over several
months. It will be used to pay down debt, to buy land, to rebuild stocks
of the product held on the farm (often a grain farmer’s savings). It will
be used to import services, especially holiday travel, and to buy assets,
a source of future income. It will be used for maintenance and the
replacement of capital equipment. In 1994, current dollar GDP in
Saskatchewan grew by 9 percent, real GDP by 5.59 percent, wages and
salaries by 2.59 percent, and total personal income fell by 0.06 percent.
Growth in the output of natural resource products does not necessarily
mean comparable rates of growth in employment either. Production from
Saskatchewan’s agriculture, and mining, quarrying, and oil wells is highly
capital intensive. There is often excess capacity, especially in the
mining sector. With the highest average weekly earnings and the second
highest hourly earnings in the mining, quarrying, and oil wells sector,
high rates of growth in exports in 1990, 1991, and 1994, for example,
meant more spending, but average monthly employment by mid-year 1996 was
still 10 percent below what it was in the mid 1980s and the growth in
monthly employment remained much below what it was in the rest of Canada.
Real gross fixed capital formation in manufacturing had fallen by 29, 58,
and 50 percent in 1992, 1993, and 1994 respectively. In public
administration it fell, on average, by half a percent each year throughout
the 1980s and 1990s. Opportunities for growth in employment had become
limited by the mid 1990s.
Will the Saskatchewan economy, output, income, and employment grow at
almost twice the Canadian rate in 1996 and 1997 as would be suggested by
the Bank of Montreal’s model if the Saskatchewan Bureau of Statistics’
market price estimates of real GDP had been used to make the forecasts? Or
will it grow at less than half the Canadian rate as would be suggested by
the Bank’s model if Statistics Canada’s market price estimates of real GDP
had been used to make the forecasts? Or at two-thirds the Canadian rate as
is suggested by the Bank’s model when the Statistics Canada factor cost
estimates of real GDP are used to make the projections?
The Bank of Montreal'’s forecasts for Saskatchewan are based o nmodest
growth in housing starts and employment with little change in the rate of
unemployment. Using its model to project the Saskatchewan Bureau’s market
price real GDP may best describe what can be expected from the growth in
provincial output in 1996 and 1997, but its forecasts of Statistics
Canada’s factor cost real GD, may best describe what can be expected of
economic activity in general: output growing at more than the Canadian
rate, even substantially more, but economic activity in general growing at
less than the Canadian rate, an ironic twist to forecasting insofar as
factor cost estimates of total real GDP refer specifically to the output
of goods and services while market price estimates are usually used to
refer to total income and spending.
By the 1970s and 1980s Saskatchewan had become a more diversified natural
resource based economy, but by the early 1990s, much of the gains from
promoting manufacturing in the 1980s had been lost. There had been little
economic development, that is, diversification which broadened the
prospects for generating employment in areas other than the export primary
product sectors. Exports of oil, potash, and uranium had become as
important as agriculture in generating incomes and employment both
directly and indirectly, but the economy remained a primary product export
economy where economic activity is excessively sensitive to the
vicissitudes of international commodity markets and the weather.
It is unusual not to use comparable estimates of output to make
projections of real GDP, as the Bank of Montreal has done, but it is
important to separate a forecast of the growth in real output from an
expectation of the prospects for economic activity. Forecasts of real GDP
are useful guidelines for indicating whether policy-makers are likely to
achieve their budget objectives. From that point of view, it may be better
to under-estimate than to over-estimate a rate of growth. Nevertheless,
forecasts that do not accurately describe what can reasonably be expected
of economic activity in the future make it impossible to judge the
effectiveness of existing policies, which new policies may be needed, and
how close budget projections will be by the end of the year.
Unfortunately, there is no way of reconciling the different rates
of growth of output which are derived from the different sets of
estimates of GDP short of making another set of estimates. There is also
no way, therefore, of reconciling what can be quite different forecasts
of economic performance. It is extremely important to supplement
forecasts of the rate of growth of real output with a wide range of
other economic statistics. Even then there is little guidance for
choosing a best forecast for any one year, even when particular models
and forecasters have been successful in the past.
Bank of Montreal, Economics Department. Regional Outlook, July,
1996. Toronto: Bank of Montreal, 1996.
Canada, Department of Finance. Economic Reference Tables, August,
1996. Catalogue no. F1-26/1996E. Ottawa: Department of Supply and
Services, Canada Communication Group, Publishing, 1996.
Canada, Statistics Canada. (11-210-XPB). Canadian Economic
Observer. Catalogue no. 11-210-XPB, 1995-96. Ottawa: Department of
Supply and Services, 1996.
Canada, Statistics Canada. (15-203-XPB). Provincial Gross Domestic
Product by Industry, 1984-1995. Catalogue no. 15-203-XPB. Ottawa:
Department of Supply and Services, 1996.
Saskatchewan, Bureau of Statistics, Saskatchewan Economic
Statistics. Regina, Bureau of Statistics, 1996.
Isabel B. Anderson, "Which Database and Which Estimates for Forecasting
Economic Performance?" Government Information in Canada/Information
gouvernementale au Canada 3, no. 1.
|