Prices of investment goods and the interest rate. Wonkish. 2 graphs.
Dear mister Eichengreen,
I read your paper ´Secular stagnation. The long view´with great interest. You discuss the secular decline of the real rate of interest, this discussion added a lot to my knowledge. That´s not just flattery. All economists should know about the developments you use to frame the discussion. You did a good job.
I am however not totally satisfied with your data on the relative price of capital goods. These can and have to be improved. This matters, as your data are already used by other researchers, see for instance ´Secular drivers of the global real interest rate´, Bank of England Staff working paper No. 571 by Lukasz Rachel and Thomas Smith and You and I know that the stakes of this kind of research are high. Very high. Conclusions should not be based upon flawed data. Which is the reason why I make the suggestions below.
The points were I see possibilities for some improvement:
- The data are too USA centered
- The source criticism is wanting
- The data are too market centered
- You use the wrong definition of savings
Let me explain. You frame the data using literature on the long-term relation between productivity, investment and the spread of different kind of innovations. I am quite enthusiastic about this discussion! When it comes to actual data on the rate of investment, which you use to show the long run development of this rate, you however only use information about the USA. Better USA data than the ones used by you are however available, as well as, more important, information about a whole bunch of other countries. Look here. These additional data in fact strengthen your case but even when they didn´t you should have taken these into account as there are clear differences in the long run pattern of the investment rate between the USA and other (western) countries. Outside the USA investment rates were often higher than in the USA. And declined much more after about somewhere between 1970 and 1990 (at least in countries without huge housing booms, where the rate of investment only declined after 2008). Which suggests that after 1970 ´something´ led to a decline in the ´strucutural´rate of investment. This ´something´ might, among other thing like the end of post WW II catching up, have been a decline of the relative price of investments. Just like lower relative prices of food contributed to a decline of the share of food in total expenditure lower prices for trucks might have led to a lower share of ´trucks´ in total expenditure. Mind however, that by far the larger amount of the total amount of real fixed capital does not consist of trucks and computers but of buildings, roads and houses.
Which leads me to the source criticism. First, you should have mentioned and discussed the definition of investments you use. As far as I could gauge you only look at Private Fixed Investments (PFI). These are defined by the Bureau of Economic Analysis as (emphasis added, it is important to note that, as national accounting is based upon a consistent stock flow logic, the classical and not the neoclassical definition of capital is used):
PFI consists of capital expenditures by private business—including expenditures on new structures, equipment, and software; net transactions in used assets; and own-account production (production by a business for its own use) of structures, equipment, and software. PFI also includes capital expenditures by nonprofit institutions serving households, and it includes capital expenditures for the acquisition of new residential structures and for improvements to existing residential structures by households in their capacity as owner-occupants. In the NIPAs, the construction of a new house (excluding the value of the unimproved land) is treated as an investment.
This leaves government investments (bridges, tanks) out of the picture. Second, a lot (in fact the larger part) of private borrowing is related to buying existing and new houses and other buildings, including, as a rule, the land underlying these buildings. Which means that the interest rate is decisively influenced by house prices, including the implied value of the land underlying houses – which however is not included in the NIPA series, which covers (among other items) construction costs of new houses excluding land. As such, the Shiller house price series might be a better metric to gauge the interest rate relevant price of investment goods than the NIPA series! Third, your graphs remind me of the early seventies (this might be a NBER thing, but that doesn´t matter). Fourth, you do not state which NIPA series you use, which matters as there are scores of NIPA series on investment prices. Some tinkering with Fred leads me to believe that you deflated the Gross Private Domestic Investment: Chain-type Price Index, Index 2009=100, Quarterly, Seasonally Adjusted with the Gross Domestic Product: Implicit Price Deflator, Index 2009=100, Quarterly, Seasonally Adjusted series (source: https://research.stlouisfed.org/fred2, assessed 21/2/2016 (if this wasn´t a blogpost but a scientific paper I should have used the original NIPA source). These series enabled me to make graph 1, very comparable to the last decades of your graph (kudos for the very long time series, of course!). The graph includes at the end two more years than your graph , which shows that the decline has at least temporarily halted (2015: 3 quarters). Scientific research has to be ´repeatable´and I seem to have managed to repeat your results. But I´m a magnitude faster than many other economists when it comes to the tinkering mentioned above. To facilitate replication and as a general rule, however, you should specify which data you used.
A problem with this series is of course that it is a USA series. One can pose the question if developments outside the USA were different. They were (graph 2). In other countries the relative price of investment also declined but less and sometimes (France) much less than in the USA. The declines also seem to have started quite a bit later. Norway even shows a large increase until about 1990, which might be connected to high investments in the oil industry. The point: it is in a scientific sense not right to use just the USA relative investment prices as an indicator which is used to explain a global decrease of real interest rates. Other data are readily available and show that sizeable differences do exist. You should have discussed this in your article (data used: Eurostat Price index (implicit deflator), 2005=100, national currency, Gross domestic product at market prices [nama_10_GDP] and Eurostat Gross fixed capital formation, price index (implicit deflator), 2010=100, national currency [nama_10_GDP], retrieved 20 February 2016).
The Eurostat series differ from your series. They are about total investments, not just private investments. But even then the conclusions holds: there is more between Los Angeles and New York (going west from LA) than just private USA investments. For the USA I checked data on relative prices for government investments. These did show some sizeable medium term differences with the relative price of private investments, but in the long run differences were small. It might however well be that these data do not encompass the prices of military equipment and I´m not completely sure if the relative price of, say, a Joint Strike Fighter also declined. Mind that in the graph it looks as if for instance German relative prices of investment goods also declined a lot but that when we restrict our attention to the period for which Eurostat data on German prices are available they declined less than half as much as USA prices. The point: it really seems that using USA prices for private investments overstates your case. Though there surely is a secular decline of prices of reproducible investment goods (but not of land, see also the Piketty data). And national differences do matter. While there are good reasons to look at government investments, including military equipment, too (in the new national accounts military equipment is treated as a capital good, which also considering the second hand market for this equipment is logical).
Which brings me to my last point. The interest rate is a nominal variable. A price which is not just set in the savings and loans market economy for reproducible investment goods (look at Spain! Ireland! Italy! Greece!). The same holds, of course, for the real interest rate. A lot of borrowing and lending is not used to finance new investment goods but to finance the purchase of existing ones, especially houses (including land underlying houses). And a lot of lending is not financed by savings but by money creation. Do not underestimate this. During the housing boom the growth rate of the stock of money in Ireland went up to about 35% a year! Or, when investments are financed by ´loanable funds borrowing´, for instance in the case of government investments in a country with a government deficit, the funds are used to buy military equipment. I know, once upon a time Tobin argued that money creating banks do not have a widows´cruce. But the combination of deregulation of the banking system with the use of ever inflating house prices to estimate the value of the collateral of mortgages made Tobin´s analysis obsolete. Money creation matters (Steve Keen and Richard Werner have written a lot about this). In the national accounts, only new investments are counted as ´savings´. And rightly so, as putting money in your mattress is, in a macro perspective, not saving. The interest rate is however not set in the ´net´lending and borrowing market but is set by gross demand and supply, including demand for loans used to buy existing assets. Interest rates are simply not only about the return on investment on new business investment goods. We´re not living in a loanable funds world.