The Level and Persistence of Growth Rates (2003)

← Back to all books

Generative illustration for The Level and Persistence of Growth Rates (2003)

Long-Term Growth Predictability

The abstract examines long-term earnings growth, which is central to valuation models and cost-of-capital estimates.
The study analyzes historical growth across a broad cross section of stocks using several operating-performance measures.
Sustained high growth is rare, and long-term earnings growth shows no persistence beyond chance.
Growth is difficult to predict: IBES forecasts are overly optimistic and add little predictive power, while valuation ratios also offer limited guidance.

Growth Expectations and Valuation

Expected growth in cash flows, commonly measured through accounting earnings, is central to stock valuation, cost-of-capital estimates, and debates over the equity risk premium.
Valuation multiples differ sharply across companies: at year-end 1999, forward price-to-earnings ratios ranged from 7.4 at the 10th percentile to 53.9 at the 90th percentile in the IBES U.S. stock universe.
Companies with sustained strong earnings growth tend to command high multiples and optimistic analyst recommendations, while past growth disappointments attract low multiples and subdued expectations.
The authors argue that risk differences are unlikely to explain most of this valuation dispersion; if prices are rational, differences in expected growth must account for much of it.
Analysts’ five-year earnings-growth forecasts also vary widely, from 8.9% at the 10th percentile to 40% at the 90th percentile, revealing a strong implicit belief that companies’ future growth can be forecast.

Other stocks with a history of dis appointing past growth are shunned by the investment community.

Questioning Persistent Earnings Growth

Market valuations and analysts’ forecasts imply that long-term earnings growth is highly predictable, but economic reasoning suggests competition should erode unusually strong growth while business exits help weaker industries recover.
The gap between optimistic expectations and actual operating performance may reflect investors’ judgmental biases or distortions in analysts’ incentives.
At year-end 1999, stocks priced at 100 times forecast earnings represented about 11.9% of total market capitalization. If that multiple fell to 20 over ten years, earnings would need to grow about 17.5% annually merely to deliver a zero return, assuming no dividends.
Under the same valuation assumptions, delivering a 10% annual stock return would require earnings growth of 29.2% annually for a decade, illustrating the demanding expectations embedded in high prices.
The paper investigates whether sustained high growth is plausible and whether firms capable of achieving it can be identified beforehand, using growth distributions over one, five, and ten years and robust persistence tests across firms, performance measures, and selected subgroups.

Following superior growth in profits, competitive pressures should ultimately tend to dilute future growth.

Earnings Growth Defies Prediction

Median operating-performance growth closely tracks GDP growth, while exceptional sustained growth is rare: only about 10 percent of firms grow faster than 18 percent annually over a decade.
Sales growth shows some persistence, but earnings growth is largely unpredictable—even among technology and pharmaceutical companies, growth stocks, and firms with strong growth histories.
Analysts’ long-term earnings-growth forecasts tend to be overly optimistic and add little predictive value over longer horizons, calling into question their widespread use in valuation and cost-of-capital estimates.
Market valuation ratios and a broader set of forecasting variables have little success distinguishing firms with high versus low future earnings growth.
The findings caution investors against paying very high valuation multiples for expected sustained growth, while extending earlier research that largely emphasized short-term earnings forecasts.

Very few firms are able to live up to the high hopes for consistent growth that are built into such rich valuations.

Measuring Corporate Growth

The study selects domestic common stocks from Compustat annually from 1951 to 1997, averaging about 2,900 firms; early data may overstate growth because of backfill bias.
Operating performance is measured using net sales, operating income before depreciation, and income before extraordinary items available to common equity.
Earnings growth is especially noisy: in an average year, 29% of firms have negative earnings, making prospective growth rates undefined, while small positive starting earnings can produce extreme growth rates.
Growth is measured per share to remove predictability attributable to changes in firm scale and to reflect the perspective of a buy-and-hold investor.
Cash dividends and special distributions are reinvested to make firms with different payout policies comparable; historical growth distributions are examined over one-, five-, and ten-year horizons.

These include such disparate cases as beaten-down companies with depressed earnings and growing startup companies that are beginning to generate profits.

Pitfalls in Measuring Growth

Reported growth rates include reinvested dividends and special distributions, making them typically higher than conventional measures; the sample’s median dividend yield averages about 2.5 percent.
Growth estimates include only firms that survive the measurement period, potentially biasing results upward as the pool shrinks substantially over longer horizons.
Income growth calculations exclude firms with negative base-year income, leaving many observations out and limiting how representative the reported distributions are.
Small positive starting incomes can produce exceptionally high growth rates: the average 98th percentile of five-year growth in income before extraordinary items is 62.4 percent annually.
Despite investors’ interest in earnings growth, its statistical behavior is especially problematic; operating income growth is less affected, while sales growth is comparatively well behaved.

Hence, while growth in income before extraordinary items captures much of the investment community’s interest, its behavior is the most questionable.

Growth Expectations and Valuation

Using sales and operating income alongside earnings provides a more robust picture of operating growth, particularly when negative starting earnings make growth rates impossible to calculate.
A stock trading at 100 times earnings would need annual earnings growth of 29.2 percent for 10 years to deliver a 10 percent annual return if its valuation multiple fell to 20—an outcome near the historical 95th percentile.
Even at a healthy annual earnings growth rate of 14.7 percent, justifying that initial multiple would require stretching the decline from 100 to 20 over 38 years.
Large firms face greater limits to growth than smaller firms: their 90th-percentile annual growth rates over 10 years are roughly 16 percent across earnings, operating income, and sales.
Median annual earnings growth over 10 years is about 10 percent across all firms, but dividend yields and inflation must also be considered when comparing this figure with real economic growth.

This is a tall order by historical standards.

Growth and Return Limits

Long-term real business growth of roughly 3.5% annually closely matches historical U.S. real GDP growth of 3.4%, suggesting corporate profitability cannot sustainably outpace the broader economy by much.
A simple dividend discount model, using historical median growth, dividend yields below 1.5%, and expected inflation of 2.5%, implies an expected stock return of only about 7.5%.
The study measures per-share growth in sales, operating income, and earnings across U.S.-listed firms from 1951 to 1998, adjusting for stock splits and reinvesting distributions.
Historical median annualized growth over ten years is approximately 10.2% for sales, 9.5% for operating income, and 9.7% for earnings, but outcomes vary substantially across firms.
Growth-rate dispersion is much wider over one year than over ten years, especially for earnings; the extreme upper tail in 1998 was also markedly higher than its historical average.

It is difficult to see how the profitability of the busi- ness sector over the long term can grow much faster than overall gross domestic product.

Testing Growth Persistence

The study measures one-, five-, and ten-year operating growth for large U.S. listed firms over 1951–1998, using per-share figures adjusted for stock splits, dividends, and distributions.
Average median annualized ten-year growth is 10.7% for sales, 9.5% for operating income before depreciation, and 9.3% for income before extraordinary items.
Growth outcomes vary much more over short horizons, especially for earnings: average one-year income growth ranges from −67.5% at the second percentile to 216.6% at the 98th percentile.
For ten-year periods ending in 1998, median growth is below the historical average for all three measures, while income growth shows more extreme outcomes at both ends of the distribution.
The next section tests the belief underlying large valuation differences: whether past growth, industry affiliation, or firm size can predict future growth. It distinguishes sustained annual growth from impressive multiyear results driven by isolated spikes.

Conversely, stocks that have done poorly over prolonged periods are shunned and trade at low multiples.

Testing Growth Persistence

The authors test growth persistence by counting “runs”: consecutive years in which a firm’s growth rate exceeds that year’s median, over horizons of one to ten years.
Each annual median includes all firms with available growth observations that year, rather than only firms surviving the full test horizon.
Survivorship bias can inflate apparent persistence: firms surviving an entire run are likely to perform better than the broader population against which they are compared.
To retain observations with negative starting earnings, the authors assign growth rates using percentile rankings of income changes scaled by stock prices; they avoid this imputation in the separate long-horizon estimates in Tables I and II.
Table III reports run counts averaged across year-ends, with the sales-growth sample beginning from an average of about 2,900 eligible firms at each year-end.

Since the survivors are likely to have better performance than the population, they tend to have a greater chance of being above the med- ian.

Persistence in Corporate Growth

The study tracks annual growth in sales, operating income, and earnings for U.S. firms listed on major exchanges from 1951 to 1998, testing runs of above-median growth over one to ten years.
Sales growth shows substantial persistence: 6.3% of eligible firms beat the median for five consecutive years, roughly twice the 3.1% expected if annual growth were independent.
Over ten years, an average of 11 surviving firms, or 0.9%, sustain above-median sales growth every year, compared with an independence benchmark of about 0.1%.
Operating income and earnings growth are much less persistent than sales growth, with consecutive above-median performance generally closer to what chance would predict.
Persistent sales growth may reflect lasting demand shifts, expansion, product improvements, or easier credit terms; it may also reflect managerial empire-building rather than profitable growth.

A firm can also sustain momentum in sales by expanding into new markets and opening new stores, by rolling out new or im- proved products, or by granting increasingly favorable credit terms.

Earnings Growth and Luck

Sales growth is easier to sustain than profit growth: discounting can boost revenue while shrinking margins, as illustrated by Internet companies whose sales rose alongside mounting losses.
Persistence in operating-income growth is weak: 3.6% of surviving firms exceed median growth for five consecutive years, only slightly above the 3.1% expected by chance.
For bottom-line earnings, sustained above-median growth occurs about as often as chance predicts, even over five- and ten-year periods.
The evidence challenges analysts’ and investors’ belief that many firms can sustain high earnings growth, and supports the view that competition erodes abnormal profits over time.
The next table examines growth persistence across technology, value, glamour, and different-sized stocks using U.S. listed-company data from 1951–1998.

Any sign of persistence vanishes as we get closer to the bottom line (Panel C).

Testing Growth Persistence

The study tests whether particular groups of firms sustain above-median growth, defining a “run” as consecutive years of growth above the median for the entire sample rather than the subgroup.
The table compares technology, value, glamour, large, mid-cap, and small stocks, using sales, operating income, and income before extraordinary items as growth measures.
Research-intensive technology and pharmaceutical firms show persistent sales and operating-income growth, but their bottom-line income growth is much closer to chance expectations.
Only 14 of 331 surviving technology stocks maintained above-median income growth for five consecutive years, compared with roughly 10 firms expected by chance; rapid sales growth need not translate into profitability.
Technology firms’ apparent growth persistence may be overstated because they are compared with slower-growing industries and because their greater volatility can intensify survivorship bias.

The recent experience of Internet companies provides numerous examples where sales grow rapidly for several years, at the same time that losses are mounting.

Growth Streaks Rarely Last

Investors may overestimate the consistency of corporate growth, extrapolating glamour stocks’ past successes and value stocks’ disappointments too far into the future.
Sales growth shows persistence in both value and glamour stocks, but other operating-performance measures reveal little difference in persistence between the two groups.
Large firms exhibit more persistent sales growth, but not income growth: only 2.2% sustain five-year above-median income-growth streaks, versus 3.2% of small firms and a chance expectation of 3.1%.
Strong past sales growth helps predict future sales growth: among surviving firms with three-year above-median sales-growth streaks, 22.8% repeat the streak over the next three years, compared with 12.5% expected by chance.
Strong past income growth offers no comparable predictive advantage: only 12.5% of surviving firms repeat a three-year above-median income-growth streak, exactly matching chance and challenging expectations embedded in growth-stock valuations.

Growth in income, on the other hand, is an entirely different matter (Panels A2 and A3).

How Growth Persists

The study examines whether firms with strong past operating growth sustain above-median growth, using U.S. listed companies over 1951–1998 and tracking performance for up to five subsequent years.
Sales growth shows the clearest persistence: among firms with five consecutive years of above-median sales growth, 63.3% outperform the median the next year, and 13.7% do so in all five subsequent years, versus 3.1% expected by chance.
Operating-income growth is less persistent than sales growth. Among firms with five-year above-median records, 8.0% sustain another five-year run; the corresponding figure for income before extraordinary items is only 5.1%.
Selecting firms by top-quartile average past growth produces a similar pattern: sales growth persists more strongly than operating-income growth, suggesting the distinction is not specific to one selection method.
Apparent excess persistence can involve very few firms: the text notes that a 5.1% repeat rate versus 3.1% expected by chance amounts to only one additional successful run, making the difference less striking than the percentages suggest.

This corresponds to only one run more than expected, however, so the difference is not outstanding.

Growth Persistence and Rebounds

The table tracks whether firms with weak growth over the preceding three or five years subsequently achieve above-median growth for one to five consecutive years.
Sales weakness tends to persist: only about 33–36% of past poor performers achieve above-median sales growth the next year, versus a 50% benchmark.
Income growth rebounds more readily than sales growth. Among firms with consecutive below-median past growth, roughly 51–58% achieve above-median operating or net income growth the following year.
Firms whose past average growth ranks in the bottom quartile show especially strong initial income rebounds: approximately 62–64% exceed median operating-income growth and 68–74% exceed median income growth before extraordinary items.
Sustained above-median growth becomes much rarer over longer horizons. The table compares these runs with a benchmark that falls from 50% for one year to 3.1% for five years, highlighting the difficulty of extrapolating growth performance.

The results caution against extrapolating past success in income growth into the

Growth Persistence Reconsidered

Ranking firms by their average growth over three or five years allows occasional weak years while still identifying strong past performers. Under this measure, past sales growth predicts future sales growth, but income growth shows no comparable persistence.
Among 103 surviving firms with top-quartile average income-growth ranks over the preceding three years, only 11 achieved above-median income growth in all three subsequent years—slightly fewer than expected by chance.
Firms with consistently weak past sales growth tend to continue struggling in sales, but their future income growth does not appear notably worse than that of past winners.
Survivorship bias complicates conclusions about past underperformers: firms that survive another observation period may disproportionately be those whose growth recovers.
Relaxing the consistency requirement to allow one or two below-median years still reveals stronger persistence in sales than in income. Sales growth beats the median in five of six years for 15.6 percent of surviving firms, versus 9.4 percent expected under independence.

Even a celebrated example of a growth stock such as Microsoft, for example, falls short of delivering above-med- ian growth in income before extraordinary items for 10 years in a row.

The High-Growth Illusion

Analysts’ forecasts and investors’ valuations suggest widespread belief that firms can sustain exceptional earnings growth, resembling belief in sporting or investing “hot hands.”
Although sales growth shows persistence, growth in operating income and earnings does not: sustained high profit growth occurs about as often as chance would predict.
The conclusion holds across firm subsets and alternative definitions of consistency, including tests that allow occasional years of weaker growth.
These findings align with economic theory: competition erodes excess earnings, causing profitability growth to revert toward normal rates and making future exceptional performers difficult to identify.
The next section raises survivorship bias as a serious concern because calculating future growth requires firms to survive, while annual comparisons include both survivors and nonsurvivors.

Put more bluntly, the chances of being able to identify the next Microsoft are about the same as the odds of winning the lottery.

Survival and Growth Predictability

The study tests survivorship bias by comparing firms that survive a full ten years with a broader group required to survive only the first five years.
Survivors are somewhat more likely to sustain above-median earnings growth: 3.4% achieve five-year runs, versus 2.3% in the broader group and a chance benchmark of 3.1%.
Survival-related differences are larger among technology stocks, where 5.2% of survivors achieve five-year growth runs compared with 3.2% of the broader group.
Survivors also have higher typical annual earnings growth: median growth averages 10.6%, compared with 8.2% for the broader group.
Although prolonged annual growth above 20% occurs, identifying such firms in advance remains difficult. The study turns to valuation ratios after finding limited predictive success using past growth, industry, investment style, and firm size.
Investors may extrapolate past growth too far into the future, potentially overvaluing past high-growth firms and undervaluing past low-growth firms.

The issue, how- ever, is whether such firms are identifiable ex ante.

Growth Persistence and Survival

The study compares firms surviving ten years with firms surviving five years but delisting before year ten, using U.S. listed-company data from 1951–1998.
Sustained above-median growth is rare, especially for earnings: only 0.2% of ten-year survivors achieve above-median income growth in every year, versus 0.9% for sales.
Survivors show greater growth persistence than nonsurvivors, while technology firms generally display stronger persistence than the broader sample, particularly in sales.
Over years six through ten, survivors have higher median annualized growth than nonsurvivors: 10.9% versus 8.4% for sales, and 10.6% versus 8.2% for income before extraordinary items.
The highest growth percentiles are similar for survivors and nonsurvivors, suggesting that eventual delisting does not rule out episodes of exceptionally strong growth.

Results for Surviving versus Non-Surviving Firms: PersistenceTests and Growth Rates

Growth and Valuation Patterns

The study groups U.S. exchange-listed firms into deciles by realized five- and ten-year earnings growth using 1951–1998 data, then compares their size and valuation ratios at the beginning and end of each horizon.
Growth outcomes vary widely: median annual earnings growth ranges from −18.9% to 41.7% across five-year deciles, versus −10.8% to 19.3% across ten-year deciles.
The fastest-growing firms tend to start smaller and gain relative size, while firms in the lowest-growth decile lose relative size over both horizons.
High-growth firms begin with lower earnings-to-price ratios, but their starting book-to-market and sales-to-price ratios are not uniformly low. Future growth therefore does not map consistently onto all initial valuation measures.
Over five years, valuation patterns diverge sharply: the lowest-growth decile’s median book-to-market ratio rises from 0.650 to 1.115, while the highest-growth decile’s falls from 0.880 to 0.560.
Firms with above-median growth in every year of a five-year period achieve median annual growth of 40.9%; their ending book-to-market ratio of 0.369 is lower than that of any growth decile.

Firms are classified into one of ten equally-sized categories based on their realized five- and ten-year growth rates.

Growth and Valuation Ambiguities

The study groups firms by realized growth and compares their size and valuation ratios at the beginning and end of five- and ten-year periods.
Firms with the most extreme growth outcomes tend to start small: the highest-growth decile grows 41.7% annually, while the lowest-growth decile declines 18.9%. Their market values subsequently rise or slump in relative terms.
Sorting firms by future growth mechanically links growth rates to starting and ending earnings. The researchers use lagged earnings to weaken this connection, but interpretation remains difficult.
A low earnings yield can reflect either expectations of strong future growth or temporarily depressed profits, making it an unreliable way to distinguish value from glamour stocks; book-to-market ratios are preferred.
Only the highest-growth decile begins with a notably low earnings yield, while other strong-growth groups do not. At the end, the lowest-growth firms also have low yields, underscoring that apparently expensive valuations need not signal strong growth.

There is reason to be wary about relying too heavily on the earnings yield vari- able, however, because net income is the most problematic of our measures of op- erating performance.

Valuations Follow Past Growth

Firms that subsequently achieve the highest earnings growth begin with relatively modest valuations, suggesting that markets fail to anticipate their success.
After strong growth occurs, firms command richer valuations, consistent with investors extrapolating past performance into future expectations.
Investors especially reward firms with five consecutive years of above-median earnings growth, apparently expecting those streaks to continue despite evidence to the contrary.
Overall, valuation ratios do little to distinguish future high-growth firms from low-growth firms; they primarily reflect growth already achieved.
The next section considers whether IBES analysts’ long-term forecasts and current dividend yields can predict earnings growth, noting analysts’ aggressive projections and the potential dilution associated with financing dividends externally.

Investors handsomely reward firms that have achieved several consecutive years of strong growth, and believe they will continue the streak (counterfactually, as the results inTableV indicate).

Analysts’ Growth Forecast Optimism

Dividend yields and future per-share growth are linked: higher current payouts can come at the expense of future growth, while payout policy may signal a firm’s investment opportunities.
To compare realized growth with analysts’ forecasts, the researchers stop assuming dividend reinvestment and examine sales and operating income alongside earnings, which present greater measurement problems.
Using 1982–1998 data, the study sorts firms into five groups by IBES consensus long-term growth forecasts and tracks surviving firms’ realized growth over one, three, and five years.
Analysts sharply distinguish growth prospects but are broadly optimistic: forecasts range from 6% in the lowest group to 22.4% in the highest, while the overall median forecast of about 14.5% exceeds realized five-year earnings growth of roughly 9%.
Forecast rankings distinguish near-term growth mainly at the extremes: first-year earnings growth averages 18.3% for the highest group versus 5.1% for the lowest, but the second- and fourth-ranked groups differ by only 2.5 percentage points.

The dispersion in IBES consensus growth forecasts is large, so analysts are boldly distinguishing between firms with high and low growth prospects.

Forecasts Versus Realized Growth

The study sorts U.S. stocks into five groups using analysts’ long-term earnings-growth forecasts and tracks subsequent operating performance over one to five years, using a 1982–1998 sample.
Higher forecast groups generally achieve higher realized sales and earnings growth, indicating that analysts’ forecasts contain useful information about relative growth prospects.
The realized growth gap is much smaller than the forecast gap: median forecasts range from 6.0% to 22.4%, while five-year annualized earnings growth for individual stocks ranges from 2.0% to 9.5%.
Value-weighted portfolio earnings growth shows a weaker, less consistent relationship with forecast rankings; over five years, the lowest- and highest-forecast groups grow by 8.0% and 11.3%, respectively.
High-forecast stocks have lower dividend yields and earnings-to-price ratios. The text introduces a naive model in which equal expected stock returns imply that differences in future growth offset differences in current dividend yields.

The naive forecast is quite successful at picking up differences in growth

Analysts Overestimate Long-Term Growth

IBES analysts’ long-term earnings-growth forecasts exceed median realized growth in every forecast quintile, with the largest optimism bias among the highest-ranked firms.
For the top quintile, median forecast growth is 22.4 percent, compared with realized five-year growth of just 9.5 percent; greater attrition in this group likely inflates even that realized figure through survivorship bias.
Analysts predict a 16.4-percentage-point growth advantage for the top quintile over the bottom, but the realized five-year gap is only 7.5 points, much of which corresponds to differences in dividend yields.
After accounting for dividend yields and survivorship bias, analyst forecasts offer additional information mainly over shorter horizons, rather than their intended three- to five-year horizon.
To address survivorship bias and problematic base-year earnings, the authors introduce value-weighted portfolio growth measures that retain all eligible firms through assumptions about delisted firms’ income and reinvestment of liquidation proceeds.

Furthermore, the realized growth rate for the firms in the top quintile should be taken with a grain of salt.

Analysts’ Growth Forecast Bias

Portfolio-based growth measures include failing firms, but reinvesting liquidation proceeds in survivors can concentrate holdings in successful companies and distort long-term results.
IBES long-term forecasts have little relationship to realized earnings growth beyond one or two years: the lowest- and highest-forecast portfolios grow by 8% and 11.3% annually over five years.
The lowest-forecast portfolio rebounds strongly in the first year, partly because firms begin with low or negative earnings and because portfolio calculations differ from individual-stock measures.
Analysts substantially overestimate growth for the top three forecast quintiles; the highest quintile’s projected growth of 22.4% compares with realized growth of roughly 11.3%–11.4%.
Forecasts are more accurate for mature, low-growth firms, while confirmation bias, enthusiasm for glamour stocks, and brokerage-related incentives may contribute to excessive optimism elsewhere.

For example, the conﬁrmation bias suggests that individuals tend to focus on evidence that supports their beliefs, while downplaying other information that is inconsistent.

Forecasting Corporate Growth

Brokerage and investment-banking incentives may encourage overly optimistic analyst forecasts, particularly in the United States, where analysts have closer ties to these businesses.
In mid-2001, analysts projected average long-term growth of about 18% for U.S. companies, compared with 11% for similarly sized companies outside the United States—a gap consistent with potential optimism bias.
The authors combine previously examined predictors into cross-sectional regression models to forecast growth in sales, operating income, and earnings over several horizons extending to five years.
The models incorporate historical sales growth, earnings yield, sustainable growth, R&D intensity, technology-sector membership, book-to-market ratios, recent stock returns, analyst forecasts, and dividend yields.
The displayed one-year results explain only about 7% of variation in sales growth and roughly 3% in operating-income and earnings growth, suggesting that even combined predictors have limited forecasting power.

We close out our analysis by gathering all the variables we have previously considered individually into one model in order to take our best shot at forecasting growth.

Testing Future Growth Predictors

The table reports regressions of future per-share growth in sales, operating income before depreciation (OIBD), and income before extraordinary items (IBEI) across several multiyear horizons.
The researchers use sales growth over the preceding five years (PASTGS5) to test whether strong past growth predicts future growth; sales growth is less erratic than earnings growth and may provide a more reliable test.
The earnings-to-price ratio (EP) is introduced as another potential predictor because it is widely interpreted as reflecting the market’s expectations of future earnings growth.
The analysis adjusts share counts for stock splits and dividends, limits extreme predictor values through Winsorization, and excludes certain firms with negative accounting measures or unusually high price-to-performance ratios.
Reported results average annual regression coefficients and model fit, with t-statistics indicating statistical significance; multiyear estimates adjust standard errors for serial correlation.

To see whether high past growth is a precursor to future growth, we use PASTGS5, the growth rate in sales over the five years prior to the sample selec- tion date.

Limits of Growth Forecasting

The forecasting model uses sustainable growth—return on equity multiplied by the earnings retention ratio—and R&D spending relative to sales to capture firms’ growth capacity and investment opportunities.
Additional predictors reflect popular signals of future growth: technology-sector membership, book-to-market ratios, recent stock returns, analysts’ long-term growth forecasts, and dividend yields.
The regression sample requires complete predictor data and positive base-year operating performance; further screens exclude negative book equity and extreme price-to-performance ratios.
Models are estimated annually, with statistical adjustments for serial correlation in regressions covering overlapping periods.
Sales growth is substantially easier to predict than bottom-line operating performance: the best model explains 11.75% of five-year sales-growth variation, while models for OIBD and IBEI barely exceed 3%. The authors caution that even these results may overstate predictability.

If anything, our results may be overstating the predictability in growth.

Limits of Growth Forecasting

Monthly regressions and alternative growth-rank tests produce similar findings, but correlations among the nine predictors make their individual contributions difficult to isolate; no variable is statistically significant across every forecasting horizon.
Past sales growth generally predicts lower future growth, suggesting reversals partly caused by depressed base-year income making subsequent growth appear stronger.
R&D intensity has the strongest forecasting power over longer horizons: spending 10 percent of sales on R&D is associated with about 2.5 percent higher five-year IBEI growth than spending nothing, although correlated predictors warrant caution.
Analysts’ long-term growth forecasts generally offer no greater predictive power than readily available dividend yields, challenging the value of expert forecasts.
The apparent link between low earnings yields and high future growth is driven largely by unusually low base-year earnings rather than a broadly reliable predictive relationship.
Technology-sector status often carries an unexpectedly negative coefficient, while book-to-market ratios and the sustainable-growth proxy do not reliably predict income growth.

In terms of predicting long-term growth, the forecasts of highly paid security analysts are about as helpful as the dividend yield, a piece of infor- mation that is readily available in the stock listings of most newspapers.

Growth Expectations Meet Reality

Historical operating-performance measures show median annual growth of about 10% with dividends reinvested during 1951–1998; excluding dividends, real growth was roughly 3–3.5%, comparable to GDP growth. Survivorship bias likely overstates expected growth.
Historical growth rates and low prevailing dividend yields imply modest prospective stock returns: approximately 7.5% under a constant-growth dividend valuation model, assuming no mispricing.
Justifying price-to-earnings ratios above 100 requires earnings growth of roughly 29% annually for at least a decade under relatively generous assumptions. Historically, only about 5% of surviving firms—and fewer than 1% of large firms—exceeded that growth rate.
Despite investors’ and analysts’ confidence that enduring high-growth firms can be identified in advance, the study finds that persistence in sales growth does not translate into persistence in earnings growth.
Simple textbook growth predictors have weak forecasting power: a model using earnings yield, sustainable growth, and R&D intensity explains only 1.48% of variation in five-year IBEI growth on an adjusted basis.

Historically, some firms have achieved such dazzling growth. These in- stances are quite rare, however.

The Growth Forecasting Illusion

Sustained earnings growth is scarcely more common than chance would predict: roughly 3% of firms achieve five consecutive years of above-median growth.
Pharmaceutical, technology, and glamour stocks offer little escape from this pattern. Sales growth shows some persistence, but strong past earnings growth does not reliably continue.
Earnings yields and other valuation ratios provide little guidance to future growth; observed valuation patterns instead support the idea that investors extrapolate past performance.
Analysts can distinguish some near-term growth prospects, but their long-term forecasts are overly optimistic and exaggerate differences between firms: a predicted growth spread of 16.4% becomes just 7.5% in realized five-year growth.
Neither analysts nor regression models offer strong long-term earnings forecasts. Analysts do little better than a simple dividend-yield benchmark, while models predict sales growth more successfully than earnings growth.

All in all, the evidence suggests that the odds of an investor successfully uncovering the next stellar growth stock are about the same as correctly calling coin tosses.

Growth Predictability and Competition

The concluding discussion emphasizes the general lack of predictability in firms’ earnings growth.
Research and development intensity stands out among the model’s variables, suggesting that intangible assets may influence future performance.
Weak growth predictability is consistent with competitive pressures correcting excessively high or low profitability growth.
The remainder of the excerpt is a reference list covering earnings forecasting, stock valuation, investor psychology, and the persistence of financial performance.

One vari- able that stands out is the level of research and development intensity, suggesting that a firm’s intangible assets may have an important influence on its future per- formance.

Finance Research References

This closing bibliography lists finance and economics research rather than presenting new arguments or findings.
Several references concern the variability of growth, including works titled “Higgledy piggledy growth” and its follow-ups.
A study on the intrinsic value of the Dow connects the references to questions of stock-market valuation.
Works on the shrinking equity premium and economists’ views of that premium highlight debate over expected stock returns.
The list includes both academic publications and a Morgan Stanley quantitative-strategy research report.

Higgledy piggledy growth in America

Kept 3 of 34 sections, 0 of 0 figures.

The High-Growth Illusion

Although sales growth shows persistence, growth in operating income and earnings does not: sustained high profit growth occurs about as often as chance would predict.
These findings align with economic theory: competition erodes excess earnings, causing profitability growth to revert toward normal rates and making future exceptional performers difficult to identify.

Put more bluntly, the chances of being able to identify the next Microsoft are about the same as the odds of winning the lottery.

Analysts Overestimate Long-Term Growth

IBES analysts’ long-term earnings-growth forecasts exceed median realized growth in every forecast quintile, with the largest optimism bias among the highest-ranked firms.
For the top quintile, median forecast growth is 22.4 percent, compared with realized five-year growth of just 9.5 percent; greater attrition in this group likely inflates even that realized figure through survivorship bias.

Furthermore, the realized growth rate for the firms in the top quintile should be taken with a grain of salt.

Growth Expectations Meet Reality

Justifying price-to-earnings ratios above 100 requires earnings growth of roughly 29% annually for at least a decade under relatively generous assumptions. Historically, only about 5% of surviving firms—and fewer than 1% of large firms—exceeded that growth rate.
Despite investors’ and analysts’ confidence that enduring high-growth firms can be identified in advance, the study finds that persistence in sales growth does not translate into persistence in earnings growth.

Historically, some firms have achieved such dazzling growth. These in- stances are quite rare, however.