2.2  The relationship between TFP and technology

Lipsey and Carlaw (2002) stated that there are three main views as to what TFP measures in the economics literature. The first is that TFP measures the rate of technological change.[5] The second is that TFP measures the “free lunches” associated with technological change, which it is argued are mainly associated with externalities and scale effects.[6] The third view is that TFP does not really measure anything useful at all.[7]

Lipsey and Carlaw argue that TFP does not measure technological change.[8] Rather, it is an imperfect measure of the “super-normal” gains that are associated with growth-creating technological change. These gains are similar, although not identical, to Jorgensen and Griliches’s (1967) concept of free lunches. They provide the following six reasons why TFP is an imperfect measure of these gains.[9]

The first is that different timings of the arrival of technological advances cause differences in measured TFP changes. Importantly these differences are more than just a different TFP time profile. When considering two different timings of a technological advance that both result in an identical final change in technology, costs and outputs, overall TFP growth may differ considerably between the two cases.

The second reason why TFP is an imperfect measure of the supernormal profits associated with technological change is due to the treatment of research and development (R&D) investment in the national accounts of some countries. In some countries R&D is recorded on the input side of the accounts as a current cost with no direct increase being made to current output. Offsetting output only appears when R&D results in reduced costs or increased final output. Such treatment means that increased R&D activity may reduce measured TFP even though there is no technological regression and even when potential technological advancement may be occurring. Their essential point is that “[i]f the patents produced are sold abroad, the transactions are recorded as capital transfers and no income is ever recorded. Hence there is no TFP gain at any point in the process.” (Carlaw and Lipsey, 2003: 19-20)

The third reason is that the omission of certain inputs, particularly natural resources, can bias measured TFP. When the quantity of the omitted variable used in the production process is growing slower than the measured inputs, TFP will be biased downwards. That is, the measured value of TFP growth will be lower than would result from a “true” measure that included accurate values for all inputs used in the production process. Conversely, if the quantity of omitted inputs is growing faster than the measured inputs, the bias will be upwards.

The fourth reason is that the use of an aggregate production function to describe disaggregated activity causes TFP growth to under-measure the free lunches associated with technological change. The reason is that if the free lunch component of the productivity advance occurs at a lower level of disaggregation than is being employed in the measurement of productivity, the free lunch will be measured as increases in inputs rather than a free increase in output for a given level of inputs.

The fifth reason is associated with the expenditure weights used to calculate the TFP index when markets are not in full equilibrium. If a technological change that confers a free lunch occurs in one sector of the economy but it takes several periods for the inputs in the economy to adjust, moving out of the relatively less productive sector into the more productive sector, then during this period of transition TFP calculated using an aggregating index such as the Divisia index will bias the measure of the free lunch depending on the returns to scale in the production function in each sector.

The sixth reason is that TFP measures the difference between contemporary changes in costs and output and consequently misses spillovers that freely provide profitable opportunities that spread across the economy as well as over long periods of time. This is yet another reason why TFP growth and technological change are not the same thing.

To illustrate the point that technological change does not necessarily show up as a change in TFP, Lipsey and Carlaw (2002) provided the following example where an upstream producer develops a piece of machinery that will allow a downstream producer to produce more output than did a previous machine. Assuming a development cost of w for the upstream firm, v representing the value of the marginal product for the downstream firm, and an initial TFP level of 1 (ie , where the output index, , equals the input index , ) , then TFP after the machine is developed is:

(1)    

and thus (as )

(2)     

This leads to the following three cases depending on the relative sizes of v and w.

1. If then
2. If then
3. If then

Although in all three cases technological change occurs, measured TFP growth can be positive, negative or zero.

While it is true that there is no one-to-one mapping between TFP and technological change, TFP measures are still widely produced and used.[10] TFP and the partial productivity measures do provide useful information on an economy’s ability to convert a given level of input into valuable outputs. Given scarcity of resources, we are interested in whether an economy can produce more output from a given level of input than was the case in an earlier period and if so by how much more.

There are several different interpretations of TFP, only one of which can be correct. For example, Bannock (1998) stated that “[i]n the long run productivity advance is the main cause of increases in real per capita income” But a number of economists would argue that productivity is actually a measured observation of the fact that increases in real per capita income have occurred. Therefore, it is not a cause but rather an observation of economic growth. This must be the case for all productivity measures. Of course if one takes productivity advance to be synonymous with technological advance then the statement makes sense. But productivity must be measured directly in this case as changes in technology, not as labour or total factor productivity.

The neoclassical growth model (Solow, 1956) points to technological progress (and by implication productivity growth) as determining long run growth in output per worker. However, it has been noted that there is good reason to believe that measured productivity does not capture technological change, and so much of technological advance that could be leading to increases in per capita output will not show up in TFP growth.[11] Thus, economists and policy makers need to clarify what productivity measures are in order to ensure that they give the correct advice about if and how to intervene in a production system in order to try to produce higher rates of sustainable economic growth.