4.2  Different approaches to constructing real GDP series (Chain versus Fixed weights) and their influence on growth rates

Annual GDP series measure the total value of goods and services produced in an economy over a 12-month period. Nominal GDP series simply sum over all possible goods and services the total value of each type of good or service produced in the 12 months. For each good or service, the total value is the number of units of the good or service produced, multiplied by the price of a unit of that good or service for that year. An increase in nominal GDP from one year to the next can therefore be attributed to an increase in prices, an increase in the volume of goods and services produced, or most probably, some combination of these two. For example, if in year 2 all prices are 10% higher than they were in year 1, and the same quantity (volume) of each good or service is produced, then nominal GDP will be 10% higher. If these goods and services must be shared amongst the same number of people in each year, is the country better off? The answer is no as in aggregate people have the same quantity of goods and services available for consumption.

Real GDP series overcome this problem by removing the impact of price changes. Consequently, changes in real GDP reflect changes in the volume or quantity of goods and services produced. Such a series is commonly referred to as being expressed in constant prices or real terms. There are several approaches to doing this, with the approaches differing in the choice of which year’s prices are used in the construction of the index.[18] As is illustrated below, the choice of which year’s prices are used has implications for the growth rates that can be obtained from the series. This section begins by considering the difference between two types of volume indexes (the Laspeyres and Paasche indexes).

A Laspeyres index calculates the total value of GDP holding prices constant at their first year levels. Table 6 presents a theoretical example of real GDP in the first and second year constructed using the Laspeyres method (note the total value of each commodity for each year utilises the first year’s prices). In the example, real GDP has grown by 18.1%.

Source: Statistics New Zealand (1998) - with very minor amendments

A Paasche index calculates the total value of GDP holding prices constant at their second (or last) year levels. Table 7 presents a theoretical example of real GDP in the first and second year using the Paasche method (note the total value of each commodity for each year utilises the last (second) year’s prices). In the example, real GDP has grown by 15.4%.

Source: Statistics New Zealand(1998) - with very minor amendments

Clearly the growth rate is dependent on which approach (Laspeyres or Paasche) is used. The result that the growth rate of the Laspeyres index is greater than the growth rate shown by the Paasche index is not just due to the construction of the example[19]. The reason why Laspeyres indexes tend to exhibit higher growth than Paasche indexes is due to the substitution effect that occurs when relative price changes occur. People tend to purchase more of goods that have become relatively cheaper and less of goods that have become relatively more expensive. Consequently goods that have become relatively cheaper tend to have faster growth (in terms of numbers of units produced and consumed) and goods that have become relatively more expensive tend to have slower growth. By using first year prices (before the relative price changes), the Laspeyres approach gives a higher weight to fast growing commodities and a smaller weight to slow growing commodities.

In terms of what drives differences in the growth rates obtained from the two approaches, Statistics New Zealand (1998) states “What matters is the extent to which the pattern of relative prices (ie the ratio of the price of one commodity to another) changes over time and not the general rate of inflation. If all prices were to increase at the same rate the two volume indices would be equal, but if some prices go up faster than others, and especially if some go down while others go up, the two volume indices will diverge. The more variation there is in the price changes, the more the volume indexes will diverge.”

Things become even more complex when one is interested in constructing values for real GDP over more than 2 periods. There are two general approaches. The first is known as the fixed weight index approach and uses the prices of just one period. Real GDP for each period in the series is calculated by multiplying the price of each commodity (in the chosen base year) by the quantity of the commodity produced in the year for which real GDP is being calculated. Until recently, this is the approach that Statistics New Zealand used and when using this approach 1991/1992 was chosen as the base year’s prices to be used. For each year, the quantity of a particular commodity produced was multiplied by that commodity’s 1991/1992 price. Summing this product over all commodities gave a value for real GDP expressed in 1991/1992 prices. Consequently, the values of real GDP prior to 1991/1992 are constructed using the Paasche index approach (as the 1991/1992 prices being used relate to a later period than the quantities of commodities produced). On the other hand, values of real GDP for years after 1991/1992 utilise the Laspeyres index approach. As a Laspeyres index tends to register higher growth rates than a Paasche index, this means that it is likely that growth prior to 1991/92 (based on a Paasche index) would be understated to growth post 1991/1992 (based on a Laspeyres index).

One issue that arises with fixed weight series is that the growth rates between consecutive years are sensitive to the choice of base year chosen. “In general, moving the base year forward in time will tend to reduce growth rates previously recorded so that they have to be revised downwards. History is rewritten.” (Statistics New Zealand, 1998) Statistics New Zealand (1998) provides an illustration of this by comparing annual growth rates for total real gross domestic expenditure when 1991/92 prices were used with the growth rates when 1982/83 prices are used.[20] Table 8 reproduces a table summarising the results. As can be seen from the table the differences in annual growth rates are quite substantial.

Source: Based on Table C from Statistics New Zealand (1998)

The second general approach to obtaining real GDP values for multiple years is known as the annual chain-linked approach and this method updates the price weights used every year. For the period 1987 to 2000, the chain-linked real GDP series is derived by calculating the (percentage) change between 1987 and 1988 using 1987 prices to value the quantities in 1987 and 1988. The change between 1988 and 1989 is calculated using 1988 prices to value the quantities in 1988 and 1989 and so on. To obtain a series of real GDP figures based on 1995 prices the following approach is used. For each year a measure of the total change between the year of interest and the year 1995 is obtained by multiplying together the annual changes between consecutive years. For years prior to 1995, the value of 1995 real GDP (which will equal the nominal GDP for 1995 as 1995 prices are being used) is divided by the by the appropriate total change figure. For years post 1995 the 1995 value for real GDP is multiplied by this amount.[21]

Note that the above approach to obtaining a chain-linked series is known as a Laspeyres chain linked approach as for each pair of years the prices of the earlier year are used. If the latest year’s prices were used for each pair the resulting index would be a Paasche chain index.

If relative prices change monotonically using chain weights instead of fixed weights tend to result in a growth rate somewhere between that of a fixed Laspeyres or fixed Paasche index. As outlined above, Statistics New Zealand has upgraded New Zealand’s National Accounts by moving from fixed to (Laspeyres) chain weights. Theoretically this should increase growth rates prior to 1991/92 as a chain Laspeyres index will produce higher growth rates than a fixed Paasche. The upgrade would also theoretically reduce growth rates after 1991/1992 as a chained Laspeyres index will result in lower growth rates than a fixed Laspeyres index.

Experimental work by Statistics New Zealand based on real (expenditure based) GDP series showed that when moving from a fixed weight method to a Laspeyres chain weighted method for constructing real GDP series the differences in (annual) growth rates are less than 0.3 percentage points although the annual growth rate between the 1994 and 1995 March years was as high as 0.6 percentage points (see Statistics New Zealand (1998) for more details).[22]

The key point to be taken from this section is that there are a number of measurement issues associated with measuring real GDP and consequently with measuring the growth in real GDP per capita. As a result there probably does not exist a definitive or ‘true’ calculated value for the historical rate of growth in a particular period. Different approaches to measuring or constructing real GDP series have resulted in the various series for real GDP per capita that are available not being identical. Therefore, the average growth rate for a period of interest will tend to vary across series.