This is work in progress. Hence the informal presentation.
Secular Stagnation
Sometime, 20-odd years back, Larry Summers alerted us to the possibility that the growth rate of affluent economies was slowing down – it was even possible it was zero for all practical purposes. Robert Gordon was another growth pessimist. Their analyses were interesting but not decisive.
I began exploring that issue for New Zealand. A paper I presented in 2011 partly reported on that exploration. I thought I had almost found the slowing down here (using the very long series of per capita GDP going back to 1858) but more intensive work concluded that I could not find it – at best it was too soon to tell.
My subsequent writing on the issue has been cautious (e.g. see Chapter 7 of In Open Seas). However, there is a persistent trope in most affluent countries’ economic debates that productivity growth is stalling. True in New Zealand, although it is not helped here by a confusion between low productivity (relative to other affluent countries) and low productivity growth. (The confusion between stock and flow – level and change – is not uncommon. It would be reduced if greater care was taken when reporting time periods for data.)
I put off further exploration because a much longer time series was needed. I have just returned to the issue using a different time series – the SNZ TFP series – which, alas, is only 35 years long. (All years reported are March years ended.)
My conclusion is that, on the whole, it seems less likely than I expected that there was evidence of a growth slowdown, but it could not be fully ruled out.
The note ends with a future possible work program arising out of this effort.
The Data Series
Statistics New Zealand publishes an annual series of volume output (GDP), and volume capital (K) and labour (L) inputs together with Total Factor Productivity (TFP), which in the Solow growth model measures the contribution of factors other than capital and labour (more below). It provides figures for individual sectors, as well as the economy as a whole.
The longer series is from (March Year ended) 1978 to 2023. There is a shorter series from 1996 to 2023 which closely replicates the longer series but has a greater decomposition of the services sector.
I followed the standard research rule of using the longer series but with the caution of taking care over the start and ending of the series. Moreover, the series is regularly revised. The 2023 version I am using reported revisions for 11 years going back to 2007. (I have not checked, but presumably they are minor ones; almost a quarter of the observations were changed. ) The 2023 estimate is ‘preliminary’.
Because this is an exploratory study, I have only investigated the aggregate GDP series. At some stage it may be worth checking the individual sectors.
The Result
The basic result is summarised in the accompanying graph. It tells the following story. Bother, the graph wont insert, although it appears when I am editing.
<><>The overall trend rate of growth of TFP was 0.94% p.a..
(I tested whether there was curvature in the linear trend by fitting a quadratic. The curvature was statistically significant, but very, very small. I am loath to rely too much on this result because it depends very heavily upon what was happening in the first six years. We need to push the series back to before 1978.)
Three phases can be discerned:
1978-1993: There was relatively slow TFP growth. I am cautious about the first few years (1978-1985), where you could argue the growth was stronger with a cycle around it, but it is the start-up period of the series; one really needs a longer series to draw a conclusion. The series certainly flattens out thereafter. This is the Rogernomics Stagnation. Perhaps there is a very mild increase towards its end. (Presumably it does not appear so much in the GDP figures because the high unemployment meant there was less labour input. This can be checked.)
1994-1998: This is a period of rapid TFP growth of 2.5% p.a. for four years. I give my interpretation below. In the interim, note that a rapid TFP growth will happen during the recovery phase of a cycle following increased utilisation by capital and labour held and retained by businesses but which they had not been able to fully deploy.
1999-2023: Over the quarter of a century, the trend TFP growth was 0.058% p.a., lower than that of the entire 1978-2023 period. Observe that response to the Asian Crisis of 1999 dropped the TFP measure, but it returned to track shortly after the cyclical recovery. That may happen after the 2022 downturn, but it is too early to tell.
Interpretation
I give two broad interpretations.
1. The long-run growth of TFP has been broadly constant and near 1.0% p.a., with a cycle (including the Rogernomics stagnation) imposed upon it.
2. The long-run growth of TFP has been constant at around 0.6% p.a. except that the ‘Rogernomics’ changes lifted TFP by a total of 4 percentage points between 1994 and 1998. However, the policy changes did not increase the subsequent growth rate.
Either interpretation leaves a mystery. We know that the GDP growth track dropped by about 15 percent between 1985 and 1995. It is hard to see from either explanation that TFP contributed much to this drop. How to explain it?
Neither explanation suggests there was a TFP growth slowdown of any magnitude (which was the original reason for the investigation).
For Further Study
1. The period before 1978 needs to be explored. I am guessing that Bryan Philpott’s work on productivity may be useful here. (Warning there is a problem with the splicing together of the GDP series before and after 1978.)
2. There is a need to reconcile the above findings with the drop in the GDP track during the Rogernomics stagnation. Extending the analysis back before 1978 is vital, but even with the current SNZ database it may be possible to make progress by taking into account the volumes of labour and capital (they are in the database). (Almost as an aside, during the period the stock of capital grew faster than GDP – so the capital to output ratio rose. What is going on here, and what are the implications?)
3. It may be worth investigating the extent to which different sectors had different TFP experiences. Additionally, it may be that a changing balance between the sectors is sufficient to affect aggregate TFP growth. (The rise of ‘low productivity’ service sectors relative to ‘high productivity’ goods sectors is a common explanation for the growth slowdown.)