It is often useful to browse through the experiences of industries which have matured ahead of our own, to get a better idea of their entrenched views of life. One industry which readily comes to mind is the consumer goods manufacturing industry.
Having spent a number of years doing advisory projects for the board of Hindustan Lever (the Unilever subsidiary in India), I learnt that a slowdown following a period of intense activity and market development is a way of life in an industry which has seen umpteen cycles. The slowdown is neither related to reaching ultimate saturation nor to the economic cycle, but they have a lot in common with the well-known multiplier-accelerator model (by Paul Samuelson) of textbook economics.
These breathers are treated as natural and inevitable rather than rude shocks, even in segments of the industry which have decades of growth potential left. For example, when evaluating a brand promotion, the natural slowdown in sales after the promotion (of course, consumers have bought forward) is anticipated, estimated and subtracted from the boom during the promotion, before calculating the net success.
The same holds true for relaunches, new advertising campaigns and the like. It is unfortunate how a similar phase in the global technology cycle, instead of being a pre-empted part of analysts' estimates, is wreaking havoc on the global financial landscape. It is not doing so by the sheer magnitude of its slowdown but by its surprise element.
The underlying principle in expecting natural breathers during innovations is simple: it doesn't matter where overall demand and supply are or where the economic cycle is. If the natural pace of sales (given consumer characteristics) was going to proceed at x% and something was done to accelerate that rate by bringing consumers forward (such as by giving a discount), there would be that many fewer new consumers left for the future.
The important aspect here is to see the acceleration relative to the natural rate x% (for the specific market and geography). It doesn't matter whether there are millions of non-users still left - if only a fraction of them were expected to enter the market in the specified timeframe, then overachieving relative to that fraction, itself, constitutes acceleration. The non-using millions will probably create a new growth story at some other time, but they would not offset a slowdown now!
It might sound extremely complicated to estimate the natural rate, but the beauty of having a framework in the background is that you might not need to do that after all (just like variables on both the left- and the right-hand side of an equation simply cancel out and are not ultimately needed).
In technology, the reason why individual waves of diffusion often follow the above pattern is explained by DeLong's Law (as Paul Krugman christened it after Brad DeLong of Berkley). Apart from the basic reason that the first part of a technology wave includes equipment sales for the installation of infrastructure in addition to actual end usage, what always works against growth is the geographical flow of adoption. For instance, take mobiles. Chart 1 shows how usership varies with per capita income (measured in purchasing power parity terms).
Chart 1: Relationship between purchasing power and mobile phone adoption
Source: ITU, World Development Report 1999/2000
The long-run growth story that people dream of is the big rise in the graph (like an S) at higher income levels. As income grows, sales explode. But in the long-run we are all dead! Readers can themselves verify that a country with per capita income of $3000 growing at 5% (per capita GDP growth rate) would take 25 years to make it to $10,000 (where the graph seems to take off).
In our relevant timeframe, the more important dynamic is about the geographical flow of adoption, which typically starts at the top end of the S-curve and gradually follows it down from right to left (see Chart 2). In short, marketers aggressively push adoption of a new technology (and consumers accept it) in richer markets first - starting with the US and Japan and ending with China and India. DCF analysis guarantees that marketers want maximum revenue upfront instead of a 20 years' growth story and if there is one tool they have to make that happen, it is the order of geographical phasing.
Chart 2: The flow of adoption works against acceleration
The other equally important dynamic is the consumer industry parallel, which can be summarized in Chart 1 by the position of an individual country relative to the natural or warranted usership level represented by the S-curve (above/below). Aggressive marketing at the cost of current profits (such as large handset subsidies) can buy consumers forward and hence take the market beyond current fundamentals (dots significantly above the S-curve). Sooner or later, though, burn rate and profit considerations reduce the marketer's aggression - consolidation and slowdown follows.
One final point on profit and burn rate considerations: that's what ultimately becomes the natural crosscheck to overachieving. Just like when a brand grows big, the ability of the marketer to lose even a small amount per unit goes down (the absolute loss starts getting too big), thereby making a discount more difficult to offer; the same applies to technology. The more one has overachieved, the more difficult it becomes to overachieve again without waiting for the past innovations to turn profitable and finance the new endeavor.
A wave of enough magnitude for the world's largest economy to outperform its trend growth calls into question the entire global savings pool, foreign capital inflow possibilities and the like. Financial considerations are much more likely to render it a constant sum game across its lifecycle than for a single market segment or brand.
From all these considerations, the final anticipated shape of the wave should be as shown below in Chart 3. The steady growth story may remain intact, but once we have outperformed the trend, the natural consequence is a compensating underperformance to bring back the average. DeLong's Law told us why the first phase should have been an outperformance.
Now the key question is how do we know when exactly outperformance ends and underperformance begins? Fortunately, simple tools can tell us. And it is here that we learn to bypass the precise quantification of the natural rate as also the need to know exactly after how much outperformance we should expect the slowdown.
Chart 3: Hypothesized volume trend
Chart 4: Growth rate and incremental growth
Think of it this way: the hypothesized shape of the wave is not ad hoc like a random walk but has a smooth cycle-like pattern. Smoothness means gradual. Theoretically, it derives from the financing considerations above, which kick in as natural brakes in a gradual manner, slowly draining marketers' ability to overachieve. Technically, smoothness precludes sudden reversals of direction. That is, once you are visibly peaking, you cannot suddenly re-accelerate - the only way forward for you is to start declining.
For such a process, incremental growth (the second derivative of sales) is a good lead indicator of growth itself. Once the wave loses incremental momentum, it peaks soon and starts declining. For a growth rate curve such as the one in Chart 4, the peak would be indicated when incremental growth goes from positive to negative (crosses zero on the right-hand axis). [Later, the same indicator would also signal the eventual end of the slowdown by crossing zero again, this time from negative into positive territory.]
This is enough indication for the directional purposes that are crucial to investment decisions. This sort of indicator merely gives us timing clues in a model we have already decided upon. It does not tell us what model is right. That, we must decide beforehand, based on parallels and hypotheses. In real life, the extraction of the correct signals from noisy data involves a bit of art as well. For one, smoothing techniques of slightly greater sophistication than moving averages are crucial to reading the structural component of data correctly (interested readers are recommended to learn the Hodrick-Prescott filter). Whenever possible, seasonally adjusted sequential growth should be used rather than year-over-year changes. If well-defined external influences or shocks are discernible in the data, these should be netted out.
In my own analysis, this type of "innovation decay" model has proved invaluable on more than one occasion. Three instances which readily spring to mind are an Indian cyclical turnaround call for end-1998 made in July 1998; the second half 2000 Asian exports slowdown call made in January 2000; and Indian consumer sector slowdown call made in August 1999. These calls are recorded in printed reports under the Dresdner Kleinwort Benson brand name (my current employer).
These cases involved very different dynamics in the common sense of the term. But they had one thing in common - all either outperformed or underperformed (as was the case with the Indian industrial cycle in 1998) their sustainable trend rates for a period because of some reason. The key analytical issue involved was to time their adjustments back to trend. Each time, a model formulated along the above lines, with some variations, followed by a careful reading of the relevant lead indicator from real life data, gave much better results than layers of linked spreadsheets.
The case of the Indian cyclical turnaround was the opposite of the current technology wave - it was already in slowdown and emerging out of it. The 1991, liberalization led to a one-time catch-up and modernization of capacities. The whole boom, as usual, was construed as a permanent increase in growth. Commensurate capacities were promptly committed. 1994-95 was the peak of that wave, after which the disappointments started. By the time the 1997-98 bottom of the cycle arrived, all kinds of structural reasons were being offered as to why India would never go anywhere.
But the above model clearly predicted that the third quarter of 1998 would be the bottom of the cycle; that there would be a vigorous recovery thereafter and that the recovery itself would fizzle out about halfway rather than become a full-fledged up cycle. Readers can now see (repeated in Chart 4b) why the recovery would have fizzled after a while - the system simply goes back to trend after the recovery.
Chart 4b: Why fizzle out?
A similar example of the Indian consumer sector involved a call for a slowdown in the growth rate for personal products (shampoo, cosmetics etc) in India. It is a fine example of how a period of aggressive penetration acceleration, relative to the rate at which it would have naturally proceeded, leads to significant periods of compensating slowdown, even when the penetration of that category is nascent and is not anywhere close to saturation.
The same framework applied to US technology investments shows how the incremental growth lead indicator indeed crossed zero during the beginning of 2000. That was the peak of the current wave. Our call that Asian exports growth would slow significantly during the second half of 2000 (made in January 2000 in Feeling guilty about feeling good) was an implication of that crossing. Chart 5 shows this lead indicator below.
Chart 5: The model applied to US IT investment growth
Source: CEIC
The interesting aspect is that the quarter-to-quarter data fluctuates around the structural trend and should continue to do so in future. This should cause alternating bouts of bearishness and bullishness. But the structural trend would be declining for a while. That has to do with the unsustainable pace and overachievement of the past and would have to wait till these excesses are purged from the system before reversing.
This structural component would not respond much to the Fed's monetary policies. Its decline has little to do with those policies. In fact, even the drastic slowdown we are currently experiencing, is the cyclical variation around the trend - I doubt whether the structural decline is yet being felt. After all, structural slowdowns are of a gradual nature rather than the grinding halt that we just went through.
The cyclical component could very well turn up with a lag following the rate cuts. The second half of this year is therefore a good time to expect better cyclical performance. But since this cyclical performance would have to take place on a declining structural trend, further disappointments are likely to abound. A final structural reversal would only be signalled by the lead indicator completing its stint in negative territory (Chart 3) and eventually turning positive. This is not a prospect anytime soon. Looks more like a cold, long, next winter than a long, hot summer now!
