Introduction
.
Over the past 30 years, numerous investing “anomalies” have been identified that cannot be explained if the markets are efficient. These factors, which provide the basis of value and growth investing, range from price multiples to momentum to financial variables that measure corporate performance. As passive investing and ETFs have grown in popularity, new smart beta products that tilt passive portfolios to these factors have taken off. However, while these factors may lead to outperformance over long periods of time, they do not work all of the time and in all areas of the market. There is still a dearth of knowledge of when, and where, these factors work, and this paper provides an answer. Like sectors that have cyclical and defensive qualities and vary in sensitivity to the economy and financial markets, factors have similar characteristics and also rotate predictably through economic and market cycles. For instance, value performs well during early and mid-expansion phases of the economy. At these times, growth is abundant, so there is no need to pay up (higher multiples) for growth; however, when growth is scarce during late economic expansions or early to mid-slowdowns, quality and growth factors outperform. Over about the last 18 months, there were four rotations between value and growth; in general, they were “on schedule” according to the model presented here. [i] Currently, the model points to growth.
.
This report is divided into a review of the data, characteristics of value and growth, the style rotation model, empirical evidence, where value screens works best, and concluding remarks.
.
Data
.
Data from FactSet consists of returns and valuation and fundamental information for US equity securities greater than or equal to $200 million market cap from 12/31/1994 through 6/1/2017. Stocks were rebalanced monthly to create quintiles of stocks with different degrees of value and growth. The valuation factor is an equal weighted composite of E/P, B/P, and S/P. Using the inverse of P/E, P/B, and P/S allowed categorization of negative earnings and book value stocks into the same quintile as high P/E and high P/B securities. Each valuation factor was converted to a percentile before the composite was created. The results were run unconstrained and with sector-neutral strategies. Outlier valuation values more than four standard deviations from the mean were removed. The returns are equal-weighted.
.
The ISM (Institute for Supply Chain Management) Manufacturing Purchasing Managers’ Index (ISM PMI) is utilized to gauge the level of economic health. This monthly index is based on equal-weighted answers to questions of purchasing and supply executives of manufacturing companies about new orders, production, employment, supplier deliveries, and inventories. A reading above or below 50 indicates the manufacturing economy is expanding or contracting over the last month. Generally, a reading above 43.3 suggests the overall economy is growing.
.
Characteristics of Growth and Value
.
Stocks can be categorized by their characteristics. The value category is simply what you pay, whereas fundamentals are what you get. Similarly, Russell ranks stocks by B/P, historical 5-year sales per share growth, and I/B/E/S forecasted 2-year medium term sales growth to determine its growth and value indices. Russell 3000 Value stocks are much cheaper on various multiples than the Russell 3000 Growth index, but the fundamentals (growth, ROE, etc.) are almost all worse for value stocks than growth (figure 1). Thus, value stocks appear to be cheap for a reason – assuming fundamentals do not improve.
.
.
Growth stocks should be priced higher than value stocks if current conditions continue. However, value managers bet on fundamentals improving and, on average, they do over time (see figures 2-3 in The Expectations Clock: A Model for Cycles and Sentiment), while growth managers hope that above average growth offsets multiple compression as high growth slows over time (see how reasonably priced quality growth stocks outperform in figure 2 of Winning with Reasonably Priced Quality Growth). Theory implies that growth stocks should trade at higher P/B, P/E, and P/S multiples than value stocks if they have above-average ROE, growth, and profit margins (figure 2). One can rearrange the Gordon Growth Model for valuation (equation on the top left) to solve for P/E, P/B, and P/S. One should be willing to pay a higher price per invested equity (P/B) if a firm earns more on its equity (ROE). Similarly, if sales are more profitable (higher net margin), then the price one is willing to pay for sales (P/S) should be higher. Finally, it is ok to pay more for earnings (P/E) if today’s earnings are low compared to the future (earnings growth is high).
.
Figures 3 and 4 provide further evidence that good companies (high ROE and EBIT margin) command higher multiples. The data is based on a sector-neutral sort of the current B/P into quintiles of all of US stocks equal to or 
greater than $200 mil in market cap. The sector neutral sort means that stocks were sorted within each sector to create medians for each quintile’s characteristics. Recall, higher ROE means there is higher earnings per level of equity investment. If income is reinvested and has the same ROE as the existing business, then high ROE companies generate high growth which further elevates the P/B multiple (see equations in figure 2). In addition, high margins may drive up ROE, and thereby growth and the P/S multiple.
.
.
The Model
.
Style rotation is predicted with four variables, two economic and two market-based. Level of economic activity and change in activity is measured by the ISM PMI. Financial market variables include whether the market is rising or falling and the magnitude of the price multiple discount of value minus growth.
.
The Economy
.
The economy rotates through cycles as corporate, consumer, and government spending fluctuate. Companies have different sensitivity to this cycle depending on their business and financial profiles. Figures 5-6 show the earnings trends before, during, and after the Financial Crisis. You can see that certain sectors (figure 5) are much more cyclical than others (figure 6). As the economy wanes, big ticket items (autos, construction) and their suppliers (steel, aluminum, and other metals) suffer. Sales decline, and these companies often have high operating and financial leverage that makes their earnings drop even further. Discretionary sales (clothing retail) may also slow or decline as unemployment rises and consumer confidence deteriorates. Finally, financial firms, which lend to the above-noted business and their customers, suffer. On the other hand, during recessions, while quality growth companies with high ROE feel the pain, their earnings and ROE barely budge in comparison (see figure 7 in Winning with Reasonably Priced Quality Growth). Health care, consumer staples, and telecom are quite stable over the various phases of the economy.
.
.
Sectors that are more cyclical are overweight in value indices versus growth indices (figure 7). Cyclical sectors tend to be more mature, so their earnings are more volatile and heavily impacted by the economy and their mature status implies they lack above average growth. Earnings volatility benefits value stocks more than growth (or stable) stocks during expansions, but amplifies problems during contractions. Low growth and volatile earnings streams means that multiples should be lower (see formulas in figure 2, higher risk and lower growth drive down the P/E).
.
.
To remove cyclical/stable sector biases, value and growth can also be analyzed within sectors. Figures 3 and 4 provide a sector neutral snapshot of value through growth stocks. The cheaper companies have worse financial results. They could be undergoing turmoil or have other problems that make them less competitive and fall behind financially versus their growth counterparts within their sectors. What they most need is a break from the “storm.” An economic expansion calms the sea after a storm. Even if cheap stocks within a sector are week versus competitors, that calming of the sea benefits them more than other stronger firms. A value stock is like a rickety old rusty tiny boat that nearly sinks during the storm, but has further room to rise as the waves calm, while a growth company is like a shining new aircraft carrier that rises high above the waves even during stormy weather. A cheap value stock’s earnings may have sunk to near zero during the recession, and the growth rate associated with rising earnings off of zero during a recovery is infinity and perhaps much higher than the highest growth rates generated by expensive quality growth companies during expansions.
.
Thee economic model for the value/growth cycle is shown in figure 8. There are six phases: (1) early expansion, (2) mid expansion, (3) late expansion, (4) early pause, (5) mid pause, and (6) late recession. The phases are measured by level of and change in the ISM PMI. The PMI is increases during the first three phases and falls in the second three phases. Level of PMI is divided as follows: < 49, 49-55 (the average PMI is 52 since 1993), and > 55. During the “risk-on” phases (dark blue, phases 1 and 2), investors rotate to value stocks which may be or are about to generate above-average earnings growth. However, as the expansion ages and late cycle arrives (phase 3), fewer and fewer companies can deliver high rates of growth. During these periods, the market may narrow, M&A and IPO activity often heats up, and a smaller group of hot growth stocks (and other growth stocks) that “excite” investors may lead the market averages higher. Then, as the economy begins to slow and growth becomes scarce (phases 4 and 5), investors are more willing to pay up (higher price multiples) for growth stocks that continue to deliver. Finally, late during a recession or slow down (phase 6), value stocks may outperform as investors forecast the next recovery. Since 1993, phases 1-3 have made up 49% of the months and phases 4-6 appear 51% of the time. Please note that the economy does not have to rotate through each of the phases in the order shown. For example, it could retreat after the PMI rises to 49-55, reaccelerate after the PMI moves down to 49-55, etc. The current phase is phase 5; the May 2017 PMI reading was 54.9 and is down over the last three months (from 56.0).
.
.
The Financial Market
.
Cheap stocks should have less room to decline during market corrections. After-all, they are already depressed. This implies that value should outperform growth during financial market corrections. Furthermore, growth stocks are high expectation stocks, so a market correction may be more of a surprise for investors in growth than in value. Value stocks are low expectation stocks and negative surprises may be more commonplace (at least in the past as the stocks were driven down). Since surprise is negative during corrections, high expectations growth stocks are often repriced down further during market corrections than already low expectation value stocks.
.
In addition, style rotations are dependent on the value spread between value and growth. That is, the difference in multiples between value and growth is sometimes wider and sometimes narrower. The P/E multiple of the Russell 3000 Value is currently 5.3 points lower than the P/E of the Russell 3000 Growth index (see last column of figure 1). This discount widens and narrows over time. Figures 9 and 10 show the EPS and return trends of large cap value and large cap growth stocks before to after the last recession. By the end of 2011, value had more than recovered all of its earnings declines during the recession and EPS was 6% higher than at the end of 2006; however, the value index was still only 75% of its 2006 year-end level. This compares to growth which was 4% higher than pre-crisis levels while EPS was 23% above. The P/E of both value and growth declined over this period, but the P/E of value (75.01/106.39 = 0.71) declined 13% more than growth (103.84/123.28 = 0.84). Assuming the outlook for value and growth are relatively stable over time and the starting P/Es were fair, this change in spread should reverse.
.
.
.
Empirical Evidence
.
Over time, on a sector-neutral or unconstrained basis, it is quite apparent that the cheapest value stocks outperform the most expensive growth stocks. Figure 11 shows the cumulative sector-neutral returns of value quintiles of stocks created from an equally-weighted composite of B/P, E/P, and S/P. The 1’s – the cheapest stocks – clearly outperform the 5’s – the most expensive stocks; however, they do not outperform all of the time (figure 12). The economic model helps explain why.
.
(Note: the dates in figure 12 refer to the beginning of the next period. For instance, the last point is May 31, 2016, and refers to the return from that date through May 31, 2017.)
.
Recall, the model has four variables. The first is change in the rate of growth, the second is growth, the third is the return of the financial market, and the fourth is the value discount of value minus growth. Let’s build the model by considering each variable successively.
.
.
Change in growth (proxy is change in the PMI) is as important, or more important, than level of growth (level of PMI). The PMI is the output of a diffusion survey (i.e., responses are increased, decreased, or remain unchanged) of month-over-month changes in manufacturing economic activity over the current month (see the FAQ page of the ISM website for more information). As business conditions change, stocks respond accordingly. Figures 13 and 14 show that 12-month change in the ISM PMI is positively correlated with the 12-month return spread between extreme value (the 1s) and extreme growth (or expensive – the 5s), and its correlation has risen to 0.82 since the financial crisis as the market has become macro driven. Similarly, the correlation between relative returns of 1s minus 3s (core/moderately valued) is also highly positively correlated with the change in the ISM PMI (not shown).
.
.
Let’s now add level of the ISM PMI to the model and narrow the time period to three months as the economy may move through phases of the economic model (figure 8) over short periods of time. For instance, from November 2014 through November 2015, the economy touched on every phase. Since that time, the environment has again moved through every phase except ISM PMI > 55 and declining (phase 4). The results of the model with change in PMI and level of PMI are shown in the first row of the table in figure 15. As predicted by the economic style model (figure 8), the outperformance of cheap stocks (1s) over expensive (3s) is highest during phases 1, 2, and 6, barely above 0% during phases 3 and 5, and negative during phase 4. These results are for sector-neutral strategies. Unconstrained (not shown), value outperforms growth in phases 1, 2, and 6, is barely above 0% in phase 5, and underperforms in phases 3 and 4.
.
.
The second and third rows in figure 15 add financial returns – the third variable – to the model. Returns are based on an equal weighting of all stocks greater than or equal to $200 million market cap. The outperformance of value over growth is greater, or the underperformance is less, in four of the six phases. Also, falling markets favor value over growth in four of the six phases (in the sector-neutral strategies shown here and for five of six phases if the strategy is unconstrained). Overall, markets rose during 64.7% of the three-month periods and declined in 35.3% of the months. In up markets, 1s outperformed 5s by 0.8%, and in the down 3-month periods they outperformed by 2.9%.
.
The final fourth piece of the puzzle that correlates with style rotation is the relative value of value versus growth. Value should be cheaper than growth, but if this discount is less than normal then, all else equal, this implies that expectations have risen for value, fallen for growth, or both. This would indicate that improvement – reversion – in value stocks is already partially priced in and the return of value less growth should be lower in the future. The opposite is also the case. If the value discount is higher than normal, then one would expect better returns of value versus growth going forward.
.
.
Similar to the multi-variate (B/P, E/P, and S/P) analysis discussed above, a single-variate analysis of just the B/P variable was performed to illustrate the efficacy of the value discount in predicting returns. B/P was chosen instead of E/P as it is generally a better indicator than E/P for predicting future returns (perhaps this is because “E” can be manipulated by management). Despite its own set of issues (historical cost, book can be written off, etc.), B/P was utilized instead of S/P since book value is less volatile than sales. B/P (or rather P/B) is also strongly positively correlated with ROE (see figure 3).
.
Figure 16 shows that relative P/B of the 1s vs 5s tends to predict changes in relative ROE. Thus, valuation predicts changes in fundamentals. Relative P/B and ROE are calculated by dividing the P/B and ROE of the 1s by the P/B and ROE of the 5s, respectively. Currently, the 1s have a P/B of 1.02, and the 5s are at 15.8, so the relative P/B is 0.065. The relative ROE is 0.284 (4.63% / 16.3%). Relative P/B has been falling over the last five or more years despite the fact that relative ROE has returned to normal levels. This implies that the market anticipates relative ROE to decline, or that value is quite cheap versus growth, or a combination of both. Figure 17 illustrates that until about 2009, when the financial markets became much more driven by macro conditions (see figure 13), changes in relative P/B were modestly negatively correlated with future 12-month returns.
.
Changes in value predict changes in fundamentals (ROE), and changes fundamentals (ROE) are positively correlated with future returns. ROE tends to rise coincident with returns; albeit, sometimes it leads (1999-2002) and lags (2009-2011) (figures 18 and 19). This implies that the size of the value discount is negatively correlated with future returns, moderated by whether the change in the value discount accurately reflects future changes in relative fundamentals (ROE) that are positively associated with coincident returns.
.
Where Valuation Works Best
.
This brings up an important point. Value is what one pays, but fundamentals are what one gets. If one gets more than one pays, then this should lead to stellar returns. Figure 20 shows that the stock returns of every value quintile are improved as ROE moves to above average from below average (1s for ROE are high ROE firms, and 5s are low ROE companies). Actually, the most expensive value quintile (5s) have some of the highest returns if those stocks also have the highest (1s) or second highest (2s) ROE. The figure also shows that moderate to high ROE firms (2s and 3s) have the best returns regardless of the valuation level (see also Winning with Reasonably Priced Quality Growth). Low ROE companies may include firms with troubles that do not adequately reverse, and the highest ROE (1s) may not be sustainable. Thus, the sweet spot – the best returns – are associated with strong but not too high of ROE. Moderation is best. As the saying goes, too much of anything can kill you!
.
.
The prior discussions show that cheap stocks (1s) tend to outperform expensive securities (5s) over time and that the efficacy varies based on the economic and financial market environment. In addition, the efficacy of value screens also varies by area of the market. Some sectors lend themselves to value investing more than growth and vice versa (see figure 21). Perhaps not surprisingly, technology, health care, and consumer discretionary stocks are the worse sectors for applying value screens. Historically, these sectors have been favored by growth managers since they, as a whole or in part (e.g., internet retail in consumer discretionary), have experienced above-average growth. The best sectors for value investing are in capital intensive mature sectors such as energy, industrials, real estate, telecommunications, and utilities. Of course, reversion – the basis of value investing – is the rule of investing law in cyclical sectors such as energy, industrials, and perhaps real estate as their earnings gyrate through economic cycles.
.
.
Final Considerations
.
Portfolios will experience periods of outperformance and underperformance simply based on their degree of exposure to value and growth. For instance, a deep value manager may underperform a value index if the economy falters simply due to being more “value” than the value index. If a manager underperforms an index, this does not necessarily mean the person is not producing alpha to his or her style. Since the economy and financial markets move through cycles, it would not be surprising to find a manager’s relative performance versus a representative index to also be cyclical.
.
It is important to understand style rotation since this will:
.
- Assist the client in evaluating fund performance;
- Assist the client in choosing value managers; and
- Improve the manager’s morale and performance.
.
It is better to evaluate fund managers based on their styles (e.g. deep value, quality value, GARP, momentum, etc.), and not some arbitrary index – even a value index for a value manager or a growth index for a growth manager – that may not fully reflect that style.
.
Understanding the forces of the economy and financial markets may help clients choose fund managers who more closely align with their strategic economic and market outlooks, and if a client does not have such an outlook, diversify among fund managers. For instance, if one expects the economy to improve, a deep value manager should be chosen; however, if one expects the economy to weaken, then a quality growth manager may be best.
.
When performance deteriorates, a portfolio manager normally faces client pressure and possibly internal pressure to improve and change. Even though portfolio performance may not be entirely driven by poor forecasts of company specific fundamentals, a portfolio manager may still question his or her positions. During these very stressful times, without a disciplined investment approach, it is not unusual for a fund manager to abandon his or her style and for clients to withdraw funds. This hurts morale, and it may also damage performance if the fund manager rotates out of positions that represent a solid approach for investing and the client withdraws funds from the portfolio just as the economy and/or markets turn and the portfolio’s stocks begin to outperform.
[i] This research is based on my experiences over the last 20+ years. I first discovered, and have been conducting research since, on cyclical factor rotation in the early 2000s before I joined a firm which was underperforming. I hypothesized that this firm’s quality-growth style was predictably out of favor during an economic expansion, which I proved in 2004 with a simple regression of its relative performance to economic variables. Some of the style rotation concepts of my past research also appear in a paper I wrote in May 2014 called “It’s a Matter of Style” for Heartland Advisors, a value-investing firm which was facing headwinds, right on schedule, as value underperformed when the economy slowed.