More On Investment: “Real” and Real

My note on the weak investment spending of Canadian businesses earlier this week sparked several comments, including one from me on the methodological problems encountered in trying to measure “real” investment effort.  Here’s some more grist for the mill of how we understand “nominal” versus “real” business investment.  Point 1 is empirical, and Point 2 is more theoretical.

1. Capital spending price indices.  Check out the 3Q 2006 national income accounts bulletin.  Tables 22 (nominal investment spending), 23 (“real” investment spending), and 30 (implicit GDP price indices) are of interest here.

By the nominal GDP table, business non-residential investment equals 11.7% of GDP, and investment in machinery & equipment equals 6.6%.  By the real table, the numbers are 13.9% and 9.2%, respectively – much stronger.  Stephen Gordon thinks the latter is a more accurate reflection of true investment, while I favour the former.

Table 30 (price indices) shows that the “price” of non-residential investment has declined just a bit (by 1 percent) since 1999.  (The indices are chained $97, but the tables in this edition of the NIEA start in 1999 – so we’ll use 1999 as our starting point.  This is convenient since 1999 was also the cyclical peak for Canadian business spending).  This 1 percent decline reflects a rough balance between a 19 percent increase in the price of non-residential structures, and a 13 percent drop in the price of M&E.  (Since M&E accounts for more than half of all non-residential spending, its price decrease slightly outweighs the larger price increase in structures – most of which, I suspect, is due to what’s happening in
Northern Alberta).  So “investment” is becoming cheaper – but not by much.

But let’s decompose these numbers a bit.  Table 30 doesn’t do this.  But we can do it ourselves, by comparing the nominal and real investment series in Tables 22 and 23.  The numbers are surprising.

As expected, the apparent “price” of computers has fallen dramatically – by 57 percent between 1999 and 2005 (and by over 70 percent since 1997, when the chain index starts).  The “price” of software has also fallen, but less dramatically so: by 5% from 1999 through 2005, and by 8% since 1997.  The “price” of all other M&E has increased over this time (consistent with the hunch I expressed in my comment on my own post).

In fact, the 1997-chained implicit price index for non-computer, non-software M&E is 123.7 in 2005.  That means that prices for this category (that we’ll call “other M&E”) have been growing considerably faster than consumer prices.

In other words, the apparent decline in the “price” of investment goods is solely due to the dramatic decline in the (quality-adjusted) “price” of computers.  Without that, the average “price” of M&E would be up by almost one-quarter since 1997.

Now, I said in my earlier comment that quality-adjusted price indices for computer products had improved from the days when they were calculated on the basis of a fixed “bundle” of goods.  (Recall, these series used to imply that the real price of computers had fallen almost to zero – because they were based on the relative computing power of a computer, using an ancient 64-KB RAM computer as the “benchmark.”)  On further reflection, however, it doesn’t seem that these indices have improved by much.  Even using a chain index, the index implies that the price of computers has fallen by 70 percent in less than a decade.  Is there any meaningful sense in which this is true?  (Remember, purchases of personal computers for business use constitute a major component of what counts as business investment in computers.)  Like most people, I suspect, I spend about the same amount of money every time I upgrade to a new generation of PC.  Yes, the computing power of that new computer is vastly improved.  But does that mean its “price” has fallen by many tens of percent?  Not at all – especially when you consider that my old generation product won’t even work on today’s software.

Businesses face the same challenge in responding to rapid technological change (more on this below).

These wild results in the M&E deflator have some equally wild consequences for estimating how much businesses are spending.  For example, in “nominal” dollars businesses spent $8.5 billion on computers in 2005.  In imaginary chain-linked 1997 “real” dollars, they spent $30 billion – almost 4 times as much.  Which is more meaningful as a measure of spending?  The $30 billion might as well be printed on Monopoly Money, for all the real economic impact that that inflated figure has.

Similarly, in nominal dollars, computers and software jointly accounted for 13% of all business M&E spending.  (That strikes me as a reasonable estimate based on what I know of real businesses.)  In imaginary chain-linked 1997 “real” dollars, however, computers and software accounted for 40% of all M&E spending.  That’s nonsensical.

Including computers and software, “real” business M&E spending has grown by over $25 billion (or almost a third) between 1999 and 2005.  Excluding computers and software, it grew by $7 billion (or 12 percent).  In fact, excluding computers and software, the growth in “real” business investment has been much weaker than the growth in measured nominal investment spending: up 17% from 1999 through 2005, versus a growth of 31% in the nominal series.  (This would be an ideal time for me to use Stephen Gordon’s argument in reverse, to claim that real investment “effort” is even weaker than the nominal data indicate.  But I won’t.  I still think it is investment spending in dollars that matters, for most macroeconomic purposes.)  Remember that corporate after-tax profits roughly doubled during this same period.

One other interesting empirical note:  The decline in the overall implicit price index for M&E was about twice as steep as the decline in the parallel implicit price deflator which StatsCan calculates for all goods imports.  Moreover, the M&E index has continued to decline through 2006 even though the dollar has started to come (somewhat) back to earth – whereas the imported goods price index has been increasing with the dollar’s softness.  From what (little) I know, I think the methodology of calculating import and export price indices is questionable.  (I’d appreciate hearing from anyone with insight on this point.)  But at any rate, it is clear that the decline in implicit M&E prices is not mostly due to currency fluctuations; it is mostly due, rather, to the bizarre impacts of quality-adjusted price indices for computer products.

It seems very clear to me from this review of current capital goods price data that the claim that “real” business investment effort is much stronger than apparent nominal spending is based completely on the strange effects of rapidly improving computer technology on deflated macroeconomic aggregates.  It is not robust or economically meaningful.

2. “Vintage” effects for business investment.  There’s a separate, conceptual point in addition to the forgoing empirical concerns.  Walrasian GE theory treats investment as the outcome of optimizing choices by individual consumers to save.  These decisions are affected by preferences (“thrift”) and by the real benefits that come with investing (“productivity”).  In this world, a decline in the prices of investment goods could indeed set off the sort of optimizing recalculation envisioned by Stephen Gordon (in his “apples versus oranges” analogy).  And with the assumption of diminishing marginal returns, it is quite reasonable to expect that a big decline in the price of investment goods would indeed only partially be reflected in more real investment, with the rest of the savings used to fund higher consumption spending (since the trade-off between current and future consumption, on which that whole loanable funds model of investment rests, has been favourably altered – the decline in real investment good prices allows the consumer to have more of both).

In reality, however, business investment decisions reflect other, more autonomous factors – in particular, strategic competition between firms (something that is assumed not to exist in Walras’ world) and ongoing technological change (something that is not captured well in Walras’ world).  Companies are forced by competition to invest continually in improved products and processes.  In this sense, they HAVE to step up their “real” investment effort (as measured by the flawed methods discussed above) or they will be driven out of business by companies which refuse to keep using 64-KB RAM computers.

In the real world, investment is not a passive consumer-led process of adding to a capital stock, in return for some confidently anticipated return (rooted in the “real productivity” of that new capital).  It is a turbulent, business-led process as companies position themselves to earn profits (or, more accurately, rents) on differentiated products or processes, anticipating the reactions of their equally-ambitious competitors.  That process of constant change is embodied in the varying vintages of concrete capital assets.

The fact that Canada’s corporations, in general (there are some exceptions, but not many), are failing to take advantage of their unique profitability to increase investment and enhance their relative competitive positions, seems to me to go a long ways toward explaining their parallel failure to succeed in differentiated global markets.  Instead, they are content to reap huge profits by digging stuff out of the ground (resources) and churn paper assets (banking).

Ironically, business investment spending is much more important to economic success in a heterodox understanding of economics, than in a neoclassical model (where investment per se carries no unique importance as a source of demand or growth).  So maybe it’s not surprising that heterodox economists should indeed be far more concerned about the weakness in Canadian business investment, than neoclassical analysts.


  • Jim, kudos for your stellar analysis of this issue.

    It is, indeed, crucial to break up aggregate “investment” into its components. Your post demonstrates that the apparent decline in the price of capital entirely reflects the supposed decline in the price of computer technology. Similarly, Joe Ruggeri and Jennifer McMullin demonstrate the Canada’s allegedly high “marginal effective tax rates” on investment are attributable to inventories, as opposed to structures or machinery and equipment:

  • I don’t have time for a proper answer (it’ll likely be a future blog post), but I really, really think you’re way off-base here. Some points:

    – The difficulties involved in computing M&E price indices are pretty much the same as those in computing the CPI. But I’m not aware of any reputable commentator who would then go on to conclude that nominal wages are the appropriate way to measure workers’ welfare.

    – The key issue here is the rate at which capital (that is, real productive capacity) is being formed: that’s what determines productivity and is the basis for any sustained increase in real wages. After a quick, 15-minute glance at the numbers, it seems to me as though estimates for the ME capital stock have been growing faster than GDP, employment and hours worked since the 2002 terms of trade shock. It’s hard to talk about falling investment when capital-output and capital-labour ratios are rising.

    That said, there are many serious questions to be raised here, such as:

    – Should investment be increasing faster?

    – Why isn’t the increase in investment showing up as increases in productivity?

  • Stephen, I am not sure that the analogy with CPI and wages supports your position.

    A criticism that I have often heard is that CPI tends to overstate inflation by not taking (sufficient) account of qualitative improvements. Jim has made the opposite criticism of the M&E deflator. In any case, computer technology (the area of contention) accounts for a far larger share of the M&E deflator than of CPI.

    To calculate labour’s share of national income, we divide total wages by GDP. (We do not deflate total wages by one index, deflate GDP by another index, and then divide “real wages” by “real GDP”.) Jim is simply suggesting that, to determine the share of national income devoted to investment, we divide total investment by GDP.

    Dropping the analogy, Stephen’s aggregate figures suggest that business is quickly expanding the capital stock. However, capital is not a homogenous thing. Jim’s (partially) disaggregated figures suggest that businesses are getting progressively better computer technology, but less other M&E and less in terms of structures. I find Jim’s story to be a more compelling description of what is happening in the Canadian economy.

  • And, crucially the political point holds: Tax relief on capital does not spur employment or wage growth but rather its opposite. And to be clear these two items are the hard core of the “popular” supply-side argument/justification for tax shifting.

    It may be that in the presence of instiutions which give workers a chance to get their share, that shifting the tax burden makes sense but in the abscence of such institutions it is a moot point.

    Alas, alas a lack.

  • For other purposes I have been doing some reading in the economics journals regarding the importance of investment spending (or not) in determining economic growth. There is quite a debate on this subject. Both neoclassical purists and some of the endogenous growth theorists prefer to credit hard-to-measure improvements in TFP as doing the main work of growth (in growth accounting studies, TFP is the “residual” — and it’s always risky to attach importance to a residual!). More pragmatic empirical studies tend to find a lot of importance for increases in the physical capital stock. Moreover, they argue that TFP improvements (to the extent they are important) are correlated with new physical investment (as opposed to coming out of thin air).

    Here are 4 recent studies I’ve found that find strong positive impacts of investment on growth: not just on income levels, but on productivity growth RATES. Interestingly (given our discussion), all these studies (as I understand them) measure investment as the nominal investment share of GDP.

    Bernanke & Gurkaynak (2001 NBER Macro Annual): “Is Growth Exogneous? Taking Mankiw, Romer, and Weil Seriously.”

    Bosworth & Collins (2003 Brookings Papers): “The Empirics of Growth: An Update.”

    D. Li (2002 Canadian Journal of Econ): “Is the AK Model Still Alive? The Long-Run Relation Between Growth and Investment Revisited.”

    Steve Bond (et al), in press somewhere (still finding the details): “Capital accumulation and growth: a new look at the evidence.”

    These citations may be useful for future work on the importance of investment spending to growth.

Leave a Reply

Your email address will not be published. Required fields are marked *