Help grow World Market Atlas — a student project for individual investors. Every contribution counts.
WORLD MARKET ATLAS · PERSONAL COMMENTARY
Market commentary
My perspective on markets, the economy and events that matter to investors.
World Market AtlasEN
S&P 500: dividends near historical lows, forward P/E across decades
Analysis: 28 Sep 2026. Indicated dividend yield: 31 Aug 2026. Complete Shiller dividends: through Jun 2026. Forward P/E: FactSet report dated 25 Sep 2026. Historical J.P. Morgan observations: 30 Mar 2026 version. Scenarios are conditional calculations, not forecasts.
The S&P 500 currently provides very little dividend income, but its forward P/E is no longer exceptionally high relative to the past decade. Both statements can be true. The official indicated dividend yield was 1.10% on 31 August 2026, while the 25 September 2026 FactSet report puts forward P/E at 19.2. One ratio measures distributions relative to price; the other measures the price of expected earnings. They answer different questions.
My assessment is that the index remains primarily an investment in future earnings development, with a small contribution from current cash distributions. A low dividend yield limits that component of returns. However, the current multiple does not support treating the market as equivalent to the peak of the dot-com bubble. The central risk is the durability of the earnings in the denominator of forward P/E. This analysis separates observations, original calculations and conditional scenarios.
1. What are we comparing?
Trailing dividend yield (TTM) divides the past 12 months of dividends by price. Indicated dividend yield uses the annualized current rate of regular distributions; it does not guarantee future income. Forward P/E divides price by consensus earnings for the next 12 months. It is distinct from P/E on reported earnings and from CAPE, which uses a long average of inflation-adjusted earnings.
Measure
Value
Period and nature of the observation
Indicated dividend yield
1.10%
S&P DJI, 31 Aug 2026
Trailing dividend yield: D/P
1.0967%
June 2026, calculated from Shiller's series
TTM dividends per index unit
USD 81.7032
Four-quarter distributions through June 2026
Forward P/E
19.2
FactSet report, 25 Sep 2026; forecast earnings
Trailing P/E
25.8
Same FactSet report; historical earnings
Sources: S&P DJI factsheet, p. 3, Robert Shiller, Data worksheet and FactSet, pp. 1 and 15. All sources were reviewed on 28 September 2026. Retrieval dates do not replace observation dates. The FactSet report uses the index closing price from 24 September; it is not an intraday reading on the date of this article.
The Shiller workbook includes prices into September, but complete dividends only through June. I do not fill in missing July–September distributions. The two yields close to 1.10% are separate observations with different definitions and dates, rather than one daily series. USD 81.70 is expressed per index unit, not per share of an arbitrary ETF.
2. Dividend history: from roughly 4% to roughly 1%
To avoid selecting individual years, I calculated statistics across every complete month in several explicit samples. Each month has equal weight. The mean is the average of monthly D/P ratios, not average dividends divided by average prices. Inputs come from Shiller's workbook.
Monthly sample
Observations
Mean yield
Median
P10–P90 range
1871–2025, broad US market history
1,860
4.22%
4.19%
1.80–6.53%
Mar 1957–Dec 2025
826
2.83%
2.89%
1.47–4.50%
1990–2025
432
1.98%
1.88%
1.31–2.89%
2016–2025
120
1.69%
1.71%
1.29–2.08%
The P10–P90 interval describes the middle 80% of observations, not a confidence interval or forecast. Data before the S&P 500's March 1957 launch represent predecessor indices and a reconstruction of the US market. They should not be described as the record of a live 500-company index operating since 1871.
The interpretation depends on the horizon. Comparing only against the 4.22% mean suggests a much larger anomaly than using the 1.69% mean from the last complete decade. Both references are valid, but they represent different market structures and distribution policies. Decade averages were 4.00% in the 1970s, 4.20% in the 1980s, 2.44% in the 1990s and 1.50% in 2020–2025. The final period contains six years rather than a complete decade.
June's 1.0967% trailing yield was about 61.3% below the March 1957–December 2025 average and 35.0% below the 2016–2025 average. It was also slightly below the August 2000 low of 1.1085% in the sample ending in 2025. Within 2026, May was marginally lower at 1.0962%. The defensible description is therefore near the historical lows of this particular series, rather than a claim that every new daily reading sets a record.
DISTRIBUTION HISTORY · MONTHLY DATA
How much dividend does the index price buy?
TTM dividends / monthly average price
Dashed line: Mar 1957–Dec 2025 average — 2.83%.Last point: June 2026, not September. August's indicated yield is a separate observation and is not appended to this series. Original calculations using Shiller data. Robert J. Shiller · ie_data.xls ↗Period averages and observation counts
Period
Months
Mean
Median
1957-03 – 1959-12
34
3.62%
3.61%
1960-01 – 1969-12
120
3.16%
3.09%
1970-01 – 1979-12
120
4.00%
3.89%
1980-01 – 1989-12
120
4.20%
4.22%
1990-01 – 1999-12
120
2.44%
2.69%
2000-01 – 2009-12
120
1.79%
1.75%
2010-01 – 2019-12
120
1.99%
1.97%
2020-01 – 2025-12
72
1.50%
1.45%
3. A low yield does not mean small nominal distributions
Separate the numerator from the denominator. Companies can pay increasing dividends while yields fall if prices rise faster. Calculations within the same series illustrate this without mixing providers.
Month
TTM dividends, USD/index unit
Monthly average index price
D/P yield
Dec 2015
43.39
2,054.08
2.11%
Dec 2020
58.28
3,695.31
1.58%
Dec 2025
79.52
6,853.03
1.16%
Jun 2026
81.70
7,450.03
1.10%
Between December 2015 and December 2025, nominal TTM dividends grew at 6.25% annually, while the index price grew at 12.80% annually. Adjusting for the change in CPI gives 2.95% annual real dividend growth. These are compound growth rates between endpoints, rather than averages of individual yearly changes. Price growth excludes reinvested dividends and therefore is not the investor's total return.
This distinction matters. Income per index unit increased, but a new investor has to pay substantially more for each unit of that income. An existing holder may receive a growing distribution while today's purchaser starts with a lower current yield. Yield on cost, calculated against an old purchase price, answers a different question from the income yield on the capital's current market value.
For a hypothetical USD 10,000 investment, a 1.10% yield corresponds to USD 110 of annual gross income if the distribution rate persists. A 2.83% yield would correspond to USD 283. This illustrates the scale of income, rather than predicting any particular fund's distribution. Fees, taxes, ETF structure and exchange rates can change the amount actually received.
4. Buybacks change the historical comparison
Dividends do not capture every transfer to shareholders. A historical example: S&P DJI's preliminary release dated 19 March 2025 reported USD 942.5 billion of buybacks and USD 629.6 billion of dividends for 2024. Buybacks therefore represented approximately 60% of the combined amount. This illustrates the distribution mix in 2024; it is not a current observation for 2026. S&P DJI.
A company repurchasing shares instead of distributing a dividend can reduce its share count and increase the remaining owners' claim on future earnings. However, gross repurchase spending is not identical to the benefit per remaining share: issuance, stock compensation and repurchase prices matter. Buying expensive shares can be poor capital allocation. Nor should a buyback yield be added to forecast EPS growth when that forecast already includes a declining share count. That would count the same contribution twice.
Within the consistent Shiller dataset, June's TTM dividends of USD 81.7032 and TTM earnings of USD 295.3881 imply a 27.7% dividend payout ratio, compared with 33.0% in December 2025. This measures dividends relative to earnings in that series, not all shareholder distributions relative to free cash flow. Multiplying August's indicated yield by September's forward P/E would not produce a precise current payout ratio: the periods and earnings definitions would be inconsistent.
Index composition matters too. The ten largest positions represented 37.8% of index weight on 31 August 2026, according to S&P DJI. In a capitalization-weighted index, a few very large companies with low distribution rates can materially reduce aggregate yield. An historical average does not automatically adjust for that change. This still does not justify any price premium: retained earnings need to be invested productively.
5. Forward P/E history does not start in 1871
There is no equally long, comparable history of analyst expectations here. I therefore do not substitute CAPE or subsequently realized earnings for historical forward P/E. The selected observations below come from J.P. Morgan Asset Management's chart in the version dated 30 March 2026. They are market turning points, not a random sample or a complete distribution of valuation multiples. Guide to the Markets, pp. 4–5.
Forward P/E was 25.2 in March 2000, 15.1 in October 2007 and 10.4 in March 2009. Before the pandemic sell-off it was 19.2 on 19 February 2020, falling to 13.3 on 23 March. The early-2022 reading was 21.4, compared with 15.7 on 12 October. Today's 19.2 therefore resembles neither dot-com euphoria nor a crisis liquidation valuation.
The 2007 comparison is particularly instructive. A moderate multiple based on expectations does not protect against a recession when those earnings subsequently disappoint. Conversely, a very high trailing P/E during a crisis may reflect a temporary collapse in reported profits. The multiple alone cannot establish whether its denominator is sustainable.
FORWARD P/E · SELECTED OBSERVATIONS
From exuberance to crisis valuations
Dot-com peak24 Mar 2000
25.2×
2002 low9 Oct 2002
14.1×
Pre-crisis peak9 Oct 2007
15.1×
Financial crisis low9 Mar 2009
10.4×
Before the pandemic19 Feb 2020
19.2×
Pandemic low23 Mar 2020
13.3×
Early 20223 Jan 2022
21.4×
2022 low12 Oct 2022
15.7×
Current FactSet report25 Sept 2026
19.2×
Common bar scale: 0–30×. Points are selected for their place in price history, not as P/E minima or maxima. They do not form a continuous series or establish percentiles. History: J.P. Morgan, 30 Mar 2026 version; final bar: FactSet, 25 Sep 2026 report. J.P. Morgan Asset Management · Guide to the Markets ↗ · FactSet · Earnings Insight ↗
6. Does 19.2 look expensive or cheap?
Reference
Forward P/E
Current 19.2 relative to the reference
FactSet five-year average, 25 Sep 2026 report
19.8
3.0% below
FactSet ten-year average, same report
19.0
1.1% above
J.P. Morgan 30-year average, as of 30 Mar 2026
17.2
11.6% above
End of Q2 2026, FactSet
20.4
5.9% below
Calculations: current multiple / reference multiple − 1. The first two comparisons use one current FactSet report; the thirty-year reference is older and comes from J.P. Morgan. It is not presented as a September recalculation. Differences in windows, consensus providers and methodology limit comparability.
The valuation is moderate against recent history but more demanding against a longer period. The five-year discount is only 3%. A small change in earnings estimates could eliminate it. Conversely, an 11.6% premium over the older thirty-year mean does not mean a fall of 11.6% must follow. Even a mechanical return to 17.2 with unchanged earnings would mean a 10.4% price decline because the denominator in that calculation differs.
An historical mean is a useful reference, not a fair-value model. Better business profitability, more stable earnings and stronger reinvestment opportunities can support a higher multiple. Higher required returns, cyclical margins and unreliable forecasts work in the opposite direction. Without assessing these factors, the phrase “above average” does not determine investment attractiveness.
7. Why did the multiple fall while the index rose?
From 30 June to the observation described in FactSet's report, the index price increased 2.7% while expected next-12-month EPS increased 8.9%. The multiple fell from 20.4 to 19.2. The relationship between price and expected earnings improved without an index price decline. FactSet, p. 15.
The identity is straightforward: P/E change = (1 + price change) / (1 + expected EPS change) − 1. The published growth rates imply approximately −5.7%; comparing the rounded multiples directly gives −5.9%. This difference reflects published precision rather than another source of return. Growth in forward EPS includes both estimate revisions and the passage of time as the twelve-month forecast window rolls forward.
In this setting, “the market became cheaper” needs a qualification: cheaper relative to forecasts. If those forecasts materialize, the improvement can be economically meaningful. If they prove too optimistic, the apparent valuation improvement can disappear. Alongside revisions, the relevant checks include revenue, margins, cash generation and financing costs.
8. Low dividends and a near-average P/E: reading them together
The inverse of 19.2 gives an earnings yield of 5.21%. This is expected earnings relative to price, not cash owed to investors. Adding the 1.10% dividend yield as a separate source of return would be incorrect because dividends distribute part of the company's resources. Likewise, simply subtracting a bond yield from earnings yield does not produce a complete model-based equity risk premium.
A low dividend yield can reflect a high price, a low dividend payout ratio, or both. Forward P/E helps assess the price channel but relies on forecasts. The current combination suggests that investment performance depends heavily on how retained profits are deployed and how the market values future earnings; a small proportion of income is delivered through current distributions.
In a simple constant-growth model, P = D₁ / (r − g), so D₁/P = r − g. A low dividend yield can be consistent with a narrow gap between the required return and long-term growth. However, this is conditional: r must exceed g, and both growth and payout policy need to be sustainable. Inserting the observed 1.10% cannot identify one reliable index forecast.
SENSITIVITY ANALYSIS · NO PROBABILITIES ASSIGNED
Two drivers of price: EPS and the multiple
Conditional scenarios, not a forecast. EPS changes versus the starting forecast, not year over year. Starting forward P/E: 19.2; price = 100. Excludes dividends, taxes, fees and currency effects.
Terminal P/E
EPS -10%
EPS 0%
EPS +10%
15.0×
-29.7%
-21.9%
-14.1%
17.2×
-19.4%
-10.4%
-1.5%
19.2×
-10.0%
0.0%
+10.0%
19.8×
-7.2%
+3.1%
+13.4%
22.0×
+3.1%
+14.6%
+26.0%
World Market Atlas calculation: (1 + ΔEPS) × terminal P/E / 19.2 − 1. A return to an historical average is a sensitivity assumption, not a prediction.
9. The cost of an earnings disappointment: sensitivity analysis
The matrix is conditional arithmetic anchored to forward P/E of 19.2. The earnings dimension represents a change relative to the forecast EPS in the starting valuation, not year-over-year earnings growth. These are neither forecasts nor probability-weighted scenarios. Price change equals (1 + EPS change) × terminal P/E / 19.2 − 1. Dividends, taxes, fees and currency effects are excluded.
The central example: with EPS 10% below the starting forecast and a multiple of 17.2, price would be 19.4% lower. Even with EPS 10% higher, a decline in the multiple to 17.2 produces a price result of roughly −1.5%. Earnings growth therefore does not guarantee positive returns if investors simultaneously pay less for each unit of profit.
There is another way to express the risk. With the price unchanged, a 10% reduction in expected EPS raises P/E from 19.2 to 21.3. The denominator can change the valuation assessment before the market price moves. A dividend yield near 1.1% is small relative to the double-digit price changes in these examples and cannot guarantee capital protection.
10. When a high dividend yield is misleading too
The opposite mistake is treating an increasing yield as automatically positive. In Shiller's series, yield rose from 1.87% in December 2007 to 3.24% in December 2008. TTM dividends then declined from USD 28.39 in December 2008 to USD 22.41 in December 2009, a 21.1% fall. A high historical distribution relative to a depressed price did not guarantee that distribution would continue. Shiller data; original calculations.
The question should therefore extend beyond the headline yield to how much earnings and cash can sustainably be distributed. During a shock, share prices can respond faster than boards change dividends. Assessing income requires cash generation, indebtedness and future investment needs. For a broad index, changing constituents and sector weights add another layer.
11. What would change the assessment?
Stronger confidence in valuation: expected profits materialize, operating cash generation improves and investment supports future profitability. An increase in consensus alone is insufficient.
A reason to reassess risk: expected EPS falls while prices hold up, margins deteriorate or repurchases weaken balance-sheet flexibility. The sensitivity analysis illustrates why the magnitude matters.
Better income quality: higher distributions supported by cash flow. A yield increase driven by collapsing share prices is not equivalent to healthier underlying businesses.
These two metrics alone do not establish an obvious bargain or an inevitable crash. Dividends are historically very low, while forward P/E is close to the past decade's average and above an older, longer-run reference. Assessing the S&P 500 consequently requires scrutiny of the quality of the earnings being purchased. Investors should not assume the dividend income of earlier decades or expect current earnings consensus to be delivered without deviations.
12. Methodology and sources
Robert J. Shiller: ie_data.xls retrieved on 28 Sep 2026; the modern D/P chart covers March 1957–June 2026. Historical comparison samples end in December 2025. Prices are monthly averages; four-quarter dividend and earnings totals are interpolated to monthly frequency. Statistics use the currently available data vintage and are not a backtest restricted to information available at each historical date. CPI is used only for real dividend growth between December 2015 and December 2025.
S&P Dow Jones Indices: factsheet dated 31 Aug 2026 for indicated yield and concentration. FactSet Earnings Insight: 25 Sep 2026 edition for forward P/E and averages, trailing P/E and changes since Q2. J.P. Morgan Asset Management: accessed version dated 30 Mar 2026 for selected historical forward P/E observations and the thirty-year mean. Publishers update the factsheet and Guide to the Markets URLs; this article's figures remain tied to the stated vintages. S&P DJI, 19 Mar 2025 release: preliminary 2024 dividends and buybacks, used as a historical example.
The dividend chart uses available monthly observations without filling gaps with zeroes. The forward P/E chart displays documented snapshots without inventing daily data between them. No forward P/E percentile or probability of a decline is estimated without a complete comparable series. Growth calculations, relative differences and sensitivity scenarios are World Market Atlas analysis, not forecasts issued by the source institutions.
Posts reflect the author’s personal views as of publication. They are not investment recommendations.
Market commentary
My perspective on markets, the economy and events that matter to investors.
S&P 500: dividends near historical lows, forward P/E across decades
Analysis: 28 Sep 2026. Indicated dividend yield: 31 Aug 2026. Complete Shiller dividends: through Jun 2026. Forward P/E: FactSet report dated 25 Sep 2026. Historical J.P. Morgan observations: 30 Mar 2026 version. Scenarios are conditional calculations, not forecasts.
The S&P 500 currently provides very little dividend income, but its forward P/E is no longer exceptionally high relative to the past decade. Both statements can be true. The official indicated dividend yield was 1.10% on 31 August 2026, while the 25 September 2026 FactSet report puts forward P/E at 19.2. One ratio measures distributions relative to price; the other measures the price of expected earnings. They answer different questions.
My assessment is that the index remains primarily an investment in future earnings development, with a small contribution from current cash distributions. A low dividend yield limits that component of returns. However, the current multiple does not support treating the market as equivalent to the peak of the dot-com bubble. The central risk is the durability of the earnings in the denominator of forward P/E. This analysis separates observations, original calculations and conditional scenarios.
1. What are we comparing?
Trailing dividend yield (TTM) divides the past 12 months of dividends by price. Indicated dividend yield uses the annualized current rate of regular distributions; it does not guarantee future income. Forward P/E divides price by consensus earnings for the next 12 months. It is distinct from P/E on reported earnings and from CAPE, which uses a long average of inflation-adjusted earnings.
Sources: S&P DJI factsheet, p. 3, Robert Shiller, Data worksheet and FactSet, pp. 1 and 15. All sources were reviewed on 28 September 2026. Retrieval dates do not replace observation dates. The FactSet report uses the index closing price from 24 September; it is not an intraday reading on the date of this article.
The Shiller workbook includes prices into September, but complete dividends only through June. I do not fill in missing July–September distributions. The two yields close to 1.10% are separate observations with different definitions and dates, rather than one daily series. USD 81.70 is expressed per index unit, not per share of an arbitrary ETF.
2. Dividend history: from roughly 4% to roughly 1%
To avoid selecting individual years, I calculated statistics across every complete month in several explicit samples. Each month has equal weight. The mean is the average of monthly D/P ratios, not average dividends divided by average prices. Inputs come from Shiller's workbook.
The P10–P90 interval describes the middle 80% of observations, not a confidence interval or forecast. Data before the S&P 500's March 1957 launch represent predecessor indices and a reconstruction of the US market. They should not be described as the record of a live 500-company index operating since 1871.
The interpretation depends on the horizon. Comparing only against the 4.22% mean suggests a much larger anomaly than using the 1.69% mean from the last complete decade. Both references are valid, but they represent different market structures and distribution policies. Decade averages were 4.00% in the 1970s, 4.20% in the 1980s, 2.44% in the 1990s and 1.50% in 2020–2025. The final period contains six years rather than a complete decade.
June's 1.0967% trailing yield was about 61.3% below the March 1957–December 2025 average and 35.0% below the 2016–2025 average. It was also slightly below the August 2000 low of 1.1085% in the sample ending in 2025. Within 2026, May was marginally lower at 1.0962%. The defensible description is therefore near the historical lows of this particular series, rather than a claim that every new daily reading sets a record.
How much dividend does the index price buy?
Period averages and observation counts
3. A low yield does not mean small nominal distributions
Separate the numerator from the denominator. Companies can pay increasing dividends while yields fall if prices rise faster. Calculations within the same series illustrate this without mixing providers.
Between December 2015 and December 2025, nominal TTM dividends grew at 6.25% annually, while the index price grew at 12.80% annually. Adjusting for the change in CPI gives 2.95% annual real dividend growth. These are compound growth rates between endpoints, rather than averages of individual yearly changes. Price growth excludes reinvested dividends and therefore is not the investor's total return.
This distinction matters. Income per index unit increased, but a new investor has to pay substantially more for each unit of that income. An existing holder may receive a growing distribution while today's purchaser starts with a lower current yield. Yield on cost, calculated against an old purchase price, answers a different question from the income yield on the capital's current market value.
For a hypothetical USD 10,000 investment, a 1.10% yield corresponds to USD 110 of annual gross income if the distribution rate persists. A 2.83% yield would correspond to USD 283. This illustrates the scale of income, rather than predicting any particular fund's distribution. Fees, taxes, ETF structure and exchange rates can change the amount actually received.
4. Buybacks change the historical comparison
Dividends do not capture every transfer to shareholders. A historical example: S&P DJI's preliminary release dated 19 March 2025 reported USD 942.5 billion of buybacks and USD 629.6 billion of dividends for 2024. Buybacks therefore represented approximately 60% of the combined amount. This illustrates the distribution mix in 2024; it is not a current observation for 2026. S&P DJI.
A company repurchasing shares instead of distributing a dividend can reduce its share count and increase the remaining owners' claim on future earnings. However, gross repurchase spending is not identical to the benefit per remaining share: issuance, stock compensation and repurchase prices matter. Buying expensive shares can be poor capital allocation. Nor should a buyback yield be added to forecast EPS growth when that forecast already includes a declining share count. That would count the same contribution twice.
Within the consistent Shiller dataset, June's TTM dividends of USD 81.7032 and TTM earnings of USD 295.3881 imply a 27.7% dividend payout ratio, compared with 33.0% in December 2025. This measures dividends relative to earnings in that series, not all shareholder distributions relative to free cash flow. Multiplying August's indicated yield by September's forward P/E would not produce a precise current payout ratio: the periods and earnings definitions would be inconsistent.
Index composition matters too. The ten largest positions represented 37.8% of index weight on 31 August 2026, according to S&P DJI. In a capitalization-weighted index, a few very large companies with low distribution rates can materially reduce aggregate yield. An historical average does not automatically adjust for that change. This still does not justify any price premium: retained earnings need to be invested productively.
5. Forward P/E history does not start in 1871
There is no equally long, comparable history of analyst expectations here. I therefore do not substitute CAPE or subsequently realized earnings for historical forward P/E. The selected observations below come from J.P. Morgan Asset Management's chart in the version dated 30 March 2026. They are market turning points, not a random sample or a complete distribution of valuation multiples. Guide to the Markets, pp. 4–5.
Forward P/E was 25.2 in March 2000, 15.1 in October 2007 and 10.4 in March 2009. Before the pandemic sell-off it was 19.2 on 19 February 2020, falling to 13.3 on 23 March. The early-2022 reading was 21.4, compared with 15.7 on 12 October. Today's 19.2 therefore resembles neither dot-com euphoria nor a crisis liquidation valuation.
The 2007 comparison is particularly instructive. A moderate multiple based on expectations does not protect against a recession when those earnings subsequently disappoint. Conversely, a very high trailing P/E during a crisis may reflect a temporary collapse in reported profits. The multiple alone cannot establish whether its denominator is sustainable.
From exuberance to crisis valuations
6. Does 19.2 look expensive or cheap?
Calculations: current multiple / reference multiple − 1. The first two comparisons use one current FactSet report; the thirty-year reference is older and comes from J.P. Morgan. It is not presented as a September recalculation. Differences in windows, consensus providers and methodology limit comparability.
The valuation is moderate against recent history but more demanding against a longer period. The five-year discount is only 3%. A small change in earnings estimates could eliminate it. Conversely, an 11.6% premium over the older thirty-year mean does not mean a fall of 11.6% must follow. Even a mechanical return to 17.2 with unchanged earnings would mean a 10.4% price decline because the denominator in that calculation differs.
An historical mean is a useful reference, not a fair-value model. Better business profitability, more stable earnings and stronger reinvestment opportunities can support a higher multiple. Higher required returns, cyclical margins and unreliable forecasts work in the opposite direction. Without assessing these factors, the phrase “above average” does not determine investment attractiveness.
7. Why did the multiple fall while the index rose?
From 30 June to the observation described in FactSet's report, the index price increased 2.7% while expected next-12-month EPS increased 8.9%. The multiple fell from 20.4 to 19.2. The relationship between price and expected earnings improved without an index price decline. FactSet, p. 15.
The identity is straightforward: P/E change = (1 + price change) / (1 + expected EPS change) − 1. The published growth rates imply approximately −5.7%; comparing the rounded multiples directly gives −5.9%. This difference reflects published precision rather than another source of return. Growth in forward EPS includes both estimate revisions and the passage of time as the twelve-month forecast window rolls forward.
In this setting, “the market became cheaper” needs a qualification: cheaper relative to forecasts. If those forecasts materialize, the improvement can be economically meaningful. If they prove too optimistic, the apparent valuation improvement can disappear. Alongside revisions, the relevant checks include revenue, margins, cash generation and financing costs.
8. Low dividends and a near-average P/E: reading them together
The inverse of 19.2 gives an earnings yield of 5.21%. This is expected earnings relative to price, not cash owed to investors. Adding the 1.10% dividend yield as a separate source of return would be incorrect because dividends distribute part of the company's resources. Likewise, simply subtracting a bond yield from earnings yield does not produce a complete model-based equity risk premium.
A low dividend yield can reflect a high price, a low dividend payout ratio, or both. Forward P/E helps assess the price channel but relies on forecasts. The current combination suggests that investment performance depends heavily on how retained profits are deployed and how the market values future earnings; a small proportion of income is delivered through current distributions.
In a simple constant-growth model, P = D₁ / (r − g), so D₁/P = r − g. A low dividend yield can be consistent with a narrow gap between the required return and long-term growth. However, this is conditional: r must exceed g, and both growth and payout policy need to be sustainable. Inserting the observed 1.10% cannot identify one reliable index forecast.
Two drivers of price: EPS and the multiple
Conditional scenarios, not a forecast. EPS changes versus the starting forecast, not year over year. Starting forward P/E: 19.2; price = 100. Excludes dividends, taxes, fees and currency effects.
World Market Atlas calculation: (1 + ΔEPS) × terminal P/E / 19.2 − 1. A return to an historical average is a sensitivity assumption, not a prediction.
9. The cost of an earnings disappointment: sensitivity analysis
The matrix is conditional arithmetic anchored to forward P/E of 19.2. The earnings dimension represents a change relative to the forecast EPS in the starting valuation, not year-over-year earnings growth. These are neither forecasts nor probability-weighted scenarios. Price change equals (1 + EPS change) × terminal P/E / 19.2 − 1. Dividends, taxes, fees and currency effects are excluded.
The central example: with EPS 10% below the starting forecast and a multiple of 17.2, price would be 19.4% lower. Even with EPS 10% higher, a decline in the multiple to 17.2 produces a price result of roughly −1.5%. Earnings growth therefore does not guarantee positive returns if investors simultaneously pay less for each unit of profit.
There is another way to express the risk. With the price unchanged, a 10% reduction in expected EPS raises P/E from 19.2 to 21.3. The denominator can change the valuation assessment before the market price moves. A dividend yield near 1.1% is small relative to the double-digit price changes in these examples and cannot guarantee capital protection.
10. When a high dividend yield is misleading too
The opposite mistake is treating an increasing yield as automatically positive. In Shiller's series, yield rose from 1.87% in December 2007 to 3.24% in December 2008. TTM dividends then declined from USD 28.39 in December 2008 to USD 22.41 in December 2009, a 21.1% fall. A high historical distribution relative to a depressed price did not guarantee that distribution would continue. Shiller data; original calculations.
The question should therefore extend beyond the headline yield to how much earnings and cash can sustainably be distributed. During a shock, share prices can respond faster than boards change dividends. Assessing income requires cash generation, indebtedness and future investment needs. For a broad index, changing constituents and sector weights add another layer.
11. What would change the assessment?
These two metrics alone do not establish an obvious bargain or an inevitable crash. Dividends are historically very low, while forward P/E is close to the past decade's average and above an older, longer-run reference. Assessing the S&P 500 consequently requires scrutiny of the quality of the earnings being purchased. Investors should not assume the dividend income of earlier decades or expect current earnings consensus to be delivered without deviations.
12. Methodology and sources
Robert J. Shiller: ie_data.xls retrieved on 28 Sep 2026; the modern D/P chart covers March 1957–June 2026. Historical comparison samples end in December 2025. Prices are monthly averages; four-quarter dividend and earnings totals are interpolated to monthly frequency. Statistics use the currently available data vintage and are not a backtest restricted to information available at each historical date. CPI is used only for real dividend growth between December 2015 and December 2025.
S&P Dow Jones Indices: factsheet dated 31 Aug 2026 for indicated yield and concentration. FactSet Earnings Insight: 25 Sep 2026 edition for forward P/E and averages, trailing P/E and changes since Q2. J.P. Morgan Asset Management: accessed version dated 30 Mar 2026 for selected historical forward P/E observations and the thirty-year mean. Publishers update the factsheet and Guide to the Markets URLs; this article's figures remain tied to the stated vintages. S&P DJI, 19 Mar 2025 release: preliminary 2024 dividends and buybacks, used as a historical example.
The dividend chart uses available monthly observations without filling gaps with zeroes. The forward P/E chart displays documented snapshots without inventing daily data between them. No forward P/E percentile or probability of a decline is estimated without a complete comparable series. Growth calculations, relative differences and sensitivity scenarios are World Market Atlas analysis, not forecasts issued by the source institutions.
Posts reflect the author’s personal views as of publication. They are not investment recommendations.