A Forecasting Model That Integrates Multiple Business and Stock Market Cycles
by Bob Bronson, Bronson Capital Markets Research. Updated July 2, 2002
This paper shows that classical business cycles can be integrated with stock-market cycles in an effective forecasting model. The Stock-Market and Economic Cycles Template (SMECT) demonstrates that these cycles are linked in a binary harmonic sequencing that is primarily trough-synchronized. Bronson's BAAC Supercycle provides the crucial missing link in this series of nested cycles Kondratiev Wave, as well as distinguishes periods of significantly higher and lower economic growth and stock market performance. SMECT shows that the interrelationships between these cycles have continued to become stronger, especially during the Modern Policymaking Era.[i] SMECT provides more refined timing for the more frequent and/or more severe bear stock markets, as well as the recessions they anticipate, that we are forecasting for the current Supercycle Bear Market Period. Finally, SMECT illustrates the applicability of Bronson's Growth Cycle to business and stock-market cycles.
Since there is an obvious interrelationship between short-term business cycles and short-term stock-market cycles, it is useful to discover their common elements in order to develop a theory that explains the underlying connections between them and, in our case, to form meaningful, differentiating forecasts, especially over longer-term horizons. By pulling back from the close-up differences and viewing the cycles from a longer-term perspective, their commonalities become more apparent - analogous to the Heisenberg Uncertainty principle, but for different feedback reasons.
It is a well-recognized fact that the stock market, as the single best leading (short-term) economic indicator, anticipates short-term business cycles. Although there have been bear markets that were not followed by recessions, during the past 75 years there has not been a U.S. recession that was not preceded by a bear market. Since 1854, there have been 32 recessions,[ii] as determined by the National Bureau of Economic Research (NBER)[iii] - each preceded by a bear stock market "anticipating" it. Most relevant for our purposes, the stock market anticipated the end of each recession with bear-market lows, or troughs, six months on average before the official end of those recessions.
Years ago, Bob Bronson, principal of Bronson Capital Markets Research, developed a useful model for predicting certain aspects of the occurrence characteristics of both stock-market and economic cycles. The template for this model graphically illustrates that the model not only explains the interrelationship of these past cycles with a high degree of accuracy - a minimum condition for any meaningful model, but it also has been, and should continue to be, a reasonably accurate forecasting tool.
The Stock Market and Business Cycles Forecasting Model
The first three charts that follow (Exhibit B) illustrate Bronson's Stock-Market and Economic Cycle Template (SMECT) for this model by incrementally overlaying several recognized business cycles. The template incorporates multiple cycles: the Kitchin[iv] cycle, the Juglar[v] cycle, the Kuznets[vi] cycle, and two so-called long-memory[vii] cycles: the Bronson Asset Allocation Cycle (BAAC) Supercycle,[viii] and the Kondratieff[ix] cycle (K-cycle).[x]
The template makes readily apparent several significant observations about the model. The model integrates and further supports the validity of the recognized Kitchin, Juglar, Kuznets, and Kondratieff business (or economic) cycles, as well as introduces their applicability to their associated stock-market cycles.[xi] The model further establishes the validity and pivotal importance of Bronson's BAAC Supercycles, which Bronson has long contended is the missing link between the Kuznets and Kondratieff cycles. Thus, this model is - as was Bronson's intention in designing it - the most comprehensive chronology of the interrelationships between classical business and stock-market cycles.
The question is, does this model help us understand stock-market and business cycles sufficiently and provide sufficient information to meaningfully forecast significant aspects of these cycles? It is clear from the template that the model not only explains historical data with a high degree of accuracy, but it also shows that the interrelationships between these cycles continue to be emergent - that is, they have become stronger. Therefore, we expect this model to continue to be a good predictor in the future - probably for at least the next 30 years - despite market efficiency and any model misspecifications,[xii] especially since it is integrated with Bronson's other submodels, factors, and indicators in his overall forecasting models.[xiii]
Based on historical facts, the theory underlying Bronson's model demonstrates that the long-memory K-cycle also is directly linked to the secular (long-term) cycles in the stock market (a subject of heated debate among some K-cycle adherents). In fact, the model explains exactly why and how the K-cycle has lengthened some 10 years[xiv] without resorting to explanations such as longer-life demographics, expanding technology-innovation eras, and other factors that may be more consequential than causal. Additionally, the template illustrates in a natural and important setting the applicability of Bronson's Growth Cycle[xv] (see Exhibit C) to stock-market and business cycles.
More importantly, historical data supports the underlying theory that these cycles are linked in a binary harmonic sequencing that is primarily trough-synchronized.[xvi] That is, the cycles can be nested one within another by lining up their troughs, with every other bottom of one cycle also the bottom of the next larger cycle, in a predictably regular sequence, or progression (see Exhibit B). The four charts in Exhibit D-1 and D-2 illustrate the historical data from the four most recent Supercycles, which cover the 28 Kitchin business cycles since 1896, including the 23 NBER-designated recessions, and the minimum set[xvii] of 31 associated bear stock markets, including the false-positives (bear markets that did not lead to recession[xviii]).
Significantly, the SMECT model also provides more refined timing for the more frequent and/or more severe bear stock markets, as well as the recessions they anticipate, that we have been forecasting for the current deflationary Supercycle bear market period since its beginning, which we called on October 7, 1997, exactly when it occurred.[xix]
Since the K-cycle does not distinguish different levels of economic growth during its uptrend (in K-cycle terminology, upgrade) and downtrend (downgrade),[xx] it serves as a useful net-cycle-neutral collective for measuring the net effects of policy-making intervention by monetary and fiscal authorities. By dividing the number of NBER recessions into the number of years in the K-cycle - which has lengthened by 10 years (endnote 14) - we can see that the NBER-designated business cycle has lengthened from about 3.4 years to about 5.0 years[xxi] since the Modern Policymaking Era began in 1913.[xxii]
The BAAC Supercycle, unlike the K-cycle, distinguishes high and low secular-growth periods in the economy. During the last three Supercycle bull market periods, there were a total of only 6 recessions whose durations aggregated only 63 months, thus averaging 10.5 months per such recession. See Exhibit A. But during the last three Supercycle bear market periods, there were a total of 13 recessions, or twice as many, and their durations aggregated to 209 months, or more than three times longer, so that their average duration was 16.1 months, or 53% longer. This is evidence of significantly higher economic growth during Supercycle bull market periods, and significantly lower economic growth during the Supercycle bear market periods.
In contrast, the 14 recessions during the upgrade phases of the K-cycle (see also Exhibit A) closely matches the 15 recessions during the downgrade phases of the last two K-cycles[xxiii] (with a Supercycle bull and a Supercycle bear market period in each K-cycle phase). This close match reflects very similar underlying economic growth during those upgrade and downgrade phases of the K-cycle.
Note that all of this is consistent with Bronson's claim that policymaking interventions, which have increased during the Modern Policymaking Era since 1913, not only have lengthened the NBER-defined business cycle, but also show up in the secular over- and underperformance of both the economy and the stock market that is reflected in BAAC Supercycle bull and bear market periods. Thus, Supercycles are important in the integration and understanding of the whole range of recognized business cycles.
The end of the previous BAAC Supercycle bull market period and the start of the current BAAC Supercycle bear market period were marked by the peak in equity asset-allocation vs. money markets and by the peaks in total return in the equally-weighted (or unweighted) and capitalization-weighted indices for all of the currently about 6,000 publicly-traded US common stocks, as follows (see also Exhibit F):
compared to money markets:
|total return (dividends reinvested) highs:||4/21/98||3/24/00|
Since10/7/97, we have been forecasting that the stock market will probably significantly underperform money markets from these two indices' respective asset-allocation peaks for about 16 (-4) years through their bear market lows at the end of the Supercycle bear market period in approximately October 2014.
We have also been forecasting that this Supercycle bear market period will probably consist of three (if one is severe) or four (regular) Kitchin business-cycle contractions, which will probably also be NBER-designated recessions (depending on the extent and effectiveness of policymaking intervention at the time), the first of which started in March 2001 and ended in November 2001. Each contraction will have a bear market anticipating it, though one or more bear markets may not be followed by recession, depending again on policymakers. The approximate timing of these business cycles is indicated in the bottom panel of the SMECT template on the second page of Exhibit D-2.
Therefore, we expect that overall economic growth probably also will be significantly lower during this Supercycle bear market period than during the preceding Supercycle bull market period (1982-97/99), which only experienced one recession from July 1990 to March 1991.
May 2002 Robert E. Bronson, III, Principal
Bronson Capital Markets Research
Anne V. Yates
[i] The Modern Policymaking Era began in 1913, the year that Congress passed legislation to establish the Federal Reserve System and added individuals to the federal income rolls (corporations were already being taxed). This new era of dramatically increased monetary and fiscal policymaking authority was first stress-tested during the Great Depression. The powers and scope of policymaking expanded rapidly in the 1930s and have grown exponentially ever since. For example, individual and corporate income taxation in relation to the total US economy has grown more than tenfold since then.
Most policymaking interventions have resulted in an increased cost of producing goods and services, which is passed on to businesses and consumers in higher prices, thereby initiating a secular rise in the rate of price inflation that arguably started from the deflationary lows of the early 1930s. However, the SMECT model considers only the effects of such interventions on the frequency and severity of business cycles and the stock-market cycles that anticipated them.
[ii] Stock-market and business cycles are economically useful because they systematically correct the inevitable, periodic accumulation of excesses in the stock market and economy. Such cycles are both self-correcting and corrected by policymakers. It can be argued that because of incomplete understanding and imperfect vision, policymaking interventions lead to inevitable unintended consequences, so that an appropriate balance between the two types of corrective forces is dynamically sought, attempting to effect maximum equity for people, businesses, and industries.
[iii] It is important to note that the NBER uses a substantially subjective definition of "recession." While two successive quarters of negative GDP is a popular definition of a recession, the NBER defines it as a period of significant decline in total output, income, employment, and trade, usually lasting from six months to a year, and marked by widespread contractions in many sectors of the economy - a definition purposely designed to have a lot of wiggle room. Not all Kitchin (see endnote 4) business-cycle contractions have led to NBER-declared recessions (see Exhibit A). Those that have come up short are usually called slowdowns, mini- or growth recessions, and/or credit crunches (e.g., 1966).
[iv] In 1923, Joseph Kitchin published in the Harvard University Press an article entitled, "Review of Economic Statistics," outlining his discovery of a primarily inventory-driven 40-month business cycle, based on his study of U.S. and U.K. financial instrument statistics from 1890-1922. The actual nomenclature "Kitchin cycle" was first used by economist Joseph A. Schumpeter (1883-1950).
[v] Clement Juglar (1891-1905), an economist considered "the father of the business cycle," studied the fluctuations in prices and interest rates in the 1860s. He determined there were boom-and-bust cycles of prosperity, crisis, liquidation, and recession. Most interesting, during the 100 years since his death, even with fiscal and monetary interventions during the Modern Policy-Making Era (footnote [i]), the eight-year Juglar cycle has persisted as demonstrated by the 13 business cycles (expansions and recessions) since then - see Exhibit A.
[vi] Economist Simon Kuznets (1901-1985) researched the U.S. infrastructure investment (including real-estate) cycle and he initiated what is now the National Income and Product Accounts, or what is commonly known as GDP data, for which he was awarded the Nobel Prize in 1971. Kuznets studied under Wesley C. Mitchell (1875-1958), who with Arthur F. Burns (1904-1987) wrote the definitive book on classical business cycles, Measuring Business Cycles , in 1946. In 1920, Mitchell founded the National Bureau of Economic Research, which has chronicled business cycle peaks and troughs since 1854.
[vii] Long memory cycles reflect long-range dependence processes that cause the cycles to be mean-reverting.
[viii] A Bronson Asset Allocation Cycle (BAAC) Supercycle is composed of one Supercycle bull followed by one Supercycle bear market period, or vice versa. A Supercycle bear market period is a 12- to 20-year period of underperformance during which bear markets, anticipating economic recessions, typically are at least twice as frequent and/or twice as severe as during Supercycle bull market periods. Such a period begins when the return on money markets sustainably exceeds the total return on equities, especially when downside-volatility-risk (DVR) is taken into account (see endnote 17). A Supercycle bull market period is a 12- to 20-year period of overperformance during which bear markets typically are only half as frequent and/or half as severe as during Supercycle bear market periods. Such a period begins when the total return on equities sustainably exceeds the return on money markets, ignoring DVR. A copy of our BAAC report, which explains this and other aspects of BAAC Supercycles, including the details of the computational methodology, may be requested by private e-mail.
[ix] In analyzing 150 years of commodity pricing history, Russian economist Nikolai D. Kondratieff (1892-1930?) discovered a supercycle of roughly three generations, now called the Kondratieff Wave or Cycle, or K-cycle. In a series of articles between 1922-28, he wrote that K-cycle theory suggested a coming depression in Western economies. Unfortunately, he angered the Communists by also correctly predicting that the cycle would be self-correcting and that the capitalist democracies would survive the Great Depression, a conclusion for which he was exiled to a gulag in 1930 - one of the few economists ever to have been exiled for their views. He was put in solitary confinement, became mentally ill and died.
The hyperinflation-deflation K-cycle has been connected with Schumpeter's technology-innovation, "creative destruction" cycle, about which we heard so much during the technology-stock bubble of the late 1990s. Today, many agree that there are four progressive economic phases of the K-cycle: K-cycle spring, reflation; K-cycle summer, inflation; K-cycle autumn - what some call the plateau, disinflation; and K-cycle winter, deflation - what some call the fall from (the) plateau. The term, deflation, is highly debated since some think it conjures economic depression or total debt liquidation, but we do not necessarily use or mean it in that way. Furthermore, some subdivide the K-cycle winter into smaller technical patterns of inflation, nominal and real interest rates, which are consistent with SMECT, but for proprietary reasons, we prefer not to address their relevance in our work at this time. Exhibit H clearly shows the K-cycle in terms of price inflation and interest rates (the inflation price of credit money).
[x] "Cycle" refers to one uptrend followed by one downtrend, or vice versa. It is important to note that these cycles are quasi-periodic cycles - not fixed time cycles. The four-year Kitchin cycle, for example, historically has varied within a range of -25% - that is, more than 90% of the time, the cycle has been no shorter than three years or longer than five years. The much longer K-cycle has only by about one-half as much variability. This decreased variability is because the nested cycles that comprise it often exhibit the self-correcting mechanism of pattern alternation, whereby shorter cycles tend to be followed by offsetting longer cycles, and vice versa, so that a series of many shorter or many longer cycles do not tend to stack up. Thus, the cumulative range (error from the mean) cuts in half. This constitutes some statistical proof that SMECT is a cohesive system of meaningfully interrelated cycles.
[xi] We attribute the same names to the stock-market cycles for ease of communication, fully recognizing that the namesake economists were only dealing with business cycles. The four-year, or Kitchin, stock market cycles nest sequentially in their aggregating Juglar, Kuznets, and Supercycle counterparts. See Exhibit B. K-cycle stock-market cycles are not meaningful in this respect.
[xii] Although SMECT was developed over many years of observation, study, and forecasting implementation, like all models, it still may not be optimally designed. Any design flaws would most likely be because of market efficiency, which results from the self-fulfilling or self-defeating feedback from competitive observers and participants in any activity whereby they can observe data, including stock-market and economic activity, and react to it, changing the outcome.
[xiii] Over the past 35 years, Bob Bronson, principal of Bronson Capital Markets Research, developed knowledge-based, "expert-system" forecasting models that quantify monetary/economic, valuation/sentiment, political/social, and inter- and intra-market technical data, adapted and reweighted in light of dynamic market changes.
[xiv] Many people think the Kondratieff cycle has an average period of about 53-54 years, but we believe that it has progressively lengthened by about 10 years to about 64 (-4) years since the start of the Modern Policymaking Era in 1913 (see endnote 21). Bronson's research and his SMECT model show that since the early 1800s, K-cycles usually have been composed of 16 Kitchin cycles (2 Kitchins/Juglar x 2 Juglars/Kuznets x 2 Kuznets/BAAC Supercycle x 2 BAAC Supercycles/K-cycle =16 Kitchins/K-cycle).
And originally Kitchin (see endnote 4) and others found the business cycle to be about 40 months long, so that the K-cycle was running about 53.3 years (= 16 x 40 / 12). But progressively over the past four to five decades, the Kitchin business cycle and the four-year presidential election cycle have morphed together, so that the Kitchin cycle has been averaging 48 months in the post-World War II period, and especially in the past four decades. Thus, the K-cycle has been progressively lengthening and is currently about 64 years (= 16 x 48 / 12), which is likely to be its final limit state because the current institutionalized form of the 48-month presidential election cycle is very likely to be very long-term stable.
[xv] Developed years ago by Bob Bronson, the Growth Cycle is a multi-fractal pattern-recognition tool that reconciles all possible chart pattern configurations. See the Growth Cycle template in Exhibit C.
[xvi] They are also peak-synchronized, but the trough synchronization is stronger. This is because the mid-term Congressional elections (early November in years 4 modulus 2 - e.g., 2002 and 2006) have become a bear-stock-market low attractor, which Bronson has demonstrated in a separate research report.
[xvii] To avoid "pick and choose" statistical bias and over-fitting the data set, the bear stock markets that anticipated these 28 Kitchin business cycles and 21 NBER-designated recessions were drawn from the smallest explanatory set of them ranked by their severity (see Exhibit E). For these purposes, severity is defined as magnitude times duration, which is an approximation of our Downside-Volatility-Risk (DVR) metric - see endnote 8. This severity approximation of DVR was used because the calculation of DVR requires daily data that is not fully available over all of the historical time periods involved in this study.
[xviii] Those bear markets did anticipate short-term, or Kitchin, business-cycle slowdowns, which did not meet the NBER's threshold for recession (see endnote 3).
[xix] At that time, we issued our lengthy report, "The Case for the Third Supercycle Bear Market Period of This Century," which may be requested by private e-mail.
[xx] It has been long recognized that the K-cycle reflects the long memory cyclical rise and fall of price inflation (goods and services) and interest rates. This can been generalized as the rise and fall of price and monetary inflation since interest rates can be regarded as the inflation rate of credit money. However, we find no satisfactory explanation in published economic literature of exactly why the K-cycle doesn't reflect secular trends in overall economic growth. That is, why is the growth rate in the economy, or aggregate production of goods and services rather than their prices, substantially similar during the K-cycle's so-called upgrade and downgrade, or up and down cycle-trends?
We offer the following explanation in frequency modulation terms: the K-cycle is the carrier frequency for its half-period subcycle, the BAAC Supercycle, which is the signal frequency, and their cycle troughs (not peaks) are synchronized. Thus, the K-cycle is a modulated carrier frequency with respect to economic growth. And while its unmodulated peak reflects the peak in price inflation and interest rates, that peak is really an economic growth trough, as illustrated by the "missing peak, double peak" chart, Exhibit G.
Here is why. Following a trough in the K-cycle, which is fully synchronized with its first four binary subcycles (Supercycle, Kuznets, Juglar and Kitchin cycles), the BAAC Supercycle and K-cycle are trending up together, and consequently, real, or inflation adjusted, economic growth is progressively higher than normal. This is the high-growth reflationary economic period, or so-called K-cycle spring. But following that spring season is the K-cycle inflationary economic summer, a Supercycle bear market period, where the carrier and signal cycles trend in opposite directions with the K-cycle continuing to trend up while the Supercycle turns and trends down. Because price inflation and interest rates are progressively rising to their eventual K-cycle peak, and are thus higher than historically normal, real economic growth progressively slows during the K-cycle summer. Consequently, at the end of the K-cycle summer and beginning of the K-cycle autumn, the turning point peak in price inflation and nominal interest rates is exactly simultaneous with the trough in progressively slowing Supercycle-measured economic growth.
Following the K-cycle summer peak, when price inflation and interest rates start progressively declining from their very high peak levels, economic growth turns around and starts expanding during that BAAC Supercycle bull market period. This is the second period where the Supercycle and K-cycle are trending in opposite directions forming the second peak illustrated in the Exhibit G "missing peak, double peak" chart.
Furthermore, this modulated carrier frequency (cycle) concept for the K-cycle explains why stock market manias occur at the midpoints of the K-cycle upgrades and downgrades - at the end of each Supercycle bull market period, or the end of the K-cycle spring and autumn economic seasons. This is illustrated for the current K-cycle since its 1949 trough in Exhibit H, along with our latest expectations. Stock market manias occur because investors become over exuberant following an extended, or Supercycle, period of economic growth where inflation and interest rates are still at moderate levels, measured on a historical basis. Of course, to those without a multi-generational long memory and only a casual understanding of these things, it seems like the business cycle has been conquered by policy-makers with no end in sight for continuing, if not escalating extant prosperity, in the then widely perceived and apparently obvious New Economic Era.
Finally, pattern alternation in these economic environments explains why every other stock market mania is a super mania with multiple technology bubbles, which always emerge at the end of K-cycle autumns or what some call plateau periods (see endnote 9), that occurred in the late-1850s, the late-1920s, and most recently, the late-1990s. The end of these disinflationary economic Supercycle bull market periods, not only occurred at moderate levels of price inflation and interest rates, but investors, consumers, businesses and policy-makers all believed that these economic indicators would only get "better" by continuing to decline - which they did, and always will - during deflationary Supercycle K-cycle winters.
Since future expectations for these economic indicators are completely distinguishable from the previous mania when price inflation and interest rates continued to rise, this is not lost upon even the most seasoned investors, consumers, businesses and policy-makers, which creates a permabull mass psychology - all believing that this (mania) time is really different. Thus, they all eagerly agree to create even more and bigger technology-based bubbles in the stock market than the previous generation did. And during the next stock market mania, the comparative logic of these competitive observer-participants reverses - and so the bigger-smaller pattern alternation in multi-generational stock market manias continues.
There are two economic reasons these every-other super manias occur, and will reoccur. First, since inflation rates are at historically moderate levels at the end of disinflationary K-cycle autumns and they continue to decline during the following K-cycle winters, there is no real capital competition from tangible (hard) capital market assets - commodities and real estate - so that trend-following, performance-oriented investors and businesses concentrate in acquiring more highly liquid, financial (intangible) capital market assets - stock and bonds. Second, because interest rates, and thus debt service cash outlays, are declining, consumers and businesses also borrow excessively, especially to fund their equity asset purchases and investments. Thus, debt-based malinvestment progressively characterizes K-cycle autumns leading to super manias and inevitable deflationary economic Supercycle bear market periods, or K-cycle winters.
[xxi] That is, 54 years/16 recessions, or 3.4 years per business cycle → 64/13, or almost 5 years per business cycle, since there were 16 NBER-designated recessions from 1858 through 1921 during the next-to-last peak-to-peak K-cycle, and only 13 such recessions during the last peak-to-peak K-cycle from 1921 through 1982 (see Exhibit A).
[xxii] The increasing intervention by fiscal and monetary policymakers - ever more frequently and more extensively - to mitigate recessions and lengthen expansions in the economy has met with increasing success, if measured over long enough periods, which we argue are minimally BAAC Supercycles. In part because of this intervention, U.S. recessions, as determined by the NBER, have become less frequent, decreasing in number from about one every less than four years in the 1800s, to about one every five years since then, if averaged over a long enough period.
While the NBER business cycle has lengthened, the Kitchin business cycle has remained about four years [see endnote 14] and, as Bronson has demonstrated elsewhere, both the Kitchin business cycle and stock market cycle are fundamentally synchronized to the four-year U.S. presidential election cycle.
However, those familiar with Bronson's research on Supercycle periods will recall that he has also determined that the frequency of recessions varies according to whether a Supercycle bull market or bear market period is underway, with recessions twice as frequent, and therefore NBER-designated business cycles half as long, during Supercycle bear market periods. Conversely, the apparent lengthening of the business cycle during the past two decades is largely explained by the less frequent recessions of a Supercycle bull market period (1982-97/99). See Exhibit A.
[xxiii] We demonstrate this model only from the K-cycle trough in 1896 because widely accepted data is only available for stock-market cycles since 1871 and for NBER-designated business cycles since 1858. However, SMECT appears to work very well going back to 1815, where we date the fourth previous K-cycle trough, rather than the 1843 date that some others use. Further, our price-inflation and interest-rate model, which covers both nominal and real interest rates, as well as the yield curve, also confirms these K-cycle peaks and troughs, as well as the Supercycle marker of the four progressive economic phases of the K-cycle (see endnote 9).
© 2002-2008 Bob Bronson
Bio and Editorial Archive