In my last post, I noted that the US has extended its dominance of global equities in recent years, increasing its share of market capitalization from 42% in at the start of 2023 to 44% at the start of 2024 to 49% at the start of 2025. That rise was driven by a surge in US equity values during 2024, with the S&P 500 delivering returns of close to 25%, all the more impressive, given that the index delivered returns in excess of 26% in 2023. In this post, I will zero in on US equities, in the aggregate, first by looking at month-by-month returns during 2024, and then putting their performance in the last two years in a historical context. I will follow up by trying to judge where markets stand at the start of 2025, starting with PE ratios, moving on to earnings yields and ending with a valuation of the index.
US Equities in 2024
Entering 2024, there was trepidation about where stocks would go during the year especially coming off a a strong bounce back year in 2023, and there remained real concerns about inflation and a recession. The hopeful note was that the Fed would lower the Fed Funds rate during the course of the year, triggering (at least in the minds of Fed watchers) lower interest rates across the yield curve, Clearly, the market not only fought through those concerns, but did so in the face of rising treasury rates, especially at the long end of the spectrum.
While the market was up strongly for the year, it is worth remembering that the there were months during 2024, where the market looked shaky, as can be seen in the month to month returns on the S&P 500 during the course of 2024:
The market’s weakest month was April 2024, and it ended the year or a weak note, down 2.50% in December. Overall, though the index was up 23.31% for the year, and adding the dividend yield of 1.57% (based upon the expected dividends for 2025 and the index at the start of the years) yields a total return 24.88% for the year:
As is almost always the case, the bulk of the returns from equity came from price appreciation, with the caveat that the dividend yield portion has shrunk over the last few decades in the United States.
Historical Context
To assess stock returns in 2024, it makes sense to step back and put the year's performance into historical perspective. In the graph below, I look at returns (inclusive of dividends) on the S&P 500 every year from 1928 to 2024.
Across the 97 years that I have estimated annual returns, stocks have had their ups and downs, delivering positive returns in 71 years and negative returns in the other 26 years. The worst year in history was 1931, with stocks returning -43.84%, and the best year was 1954, when the annual return was 52.56%. If you wanted to pick a benchmark to compare annual returns to pass judgment on whether a year was above or below average, you can can go with either the annual return (11.79%) or the median return (14.82%) across the entire time period.
Looking at the 24.88% return in 2024 in terms of rankings, it ranks as the 27th best year across the last 97 years, indicating that while it was a good year, there have been far better years for US stocks. Combining 2023 and 2024 returns yield a cumulative a two-year return for the S&P 500 of 57.42%, making it one the ten best two-year periods in US market history.
The riskless alternative to investing in US stocks during this period, in US dollar terms, are US treasuries, and in 2024, that contest was won, hands down, by US equities:
Equity risk premium earned in 2024, over 3-month treasury bills
= Return on stocks - Return on 3-month treasuries (averaged over 2024)
= 24.88% -4.97% = 19.91%
Equity risk premium earned in 2024, over 10-year treasuries
= Return on stocks - Return on 10-year treasury
= 24.88% -(-1.64%) = 26.52%
The ten-year treasury return was negative, because treasury bond rates rose during 2024.
Equity risk premiums are volatile over time, and averaging them makes sense, and in the table below, I look at the premium that stocks have earned over treasury bills and treasury bonds, going back to 1928, using both simple averages (of the returns each year) and geometric averages (reflecting the compounding effect):
These returns are nominal returns, and inflation would have taken a bite out of returns each year. Computing the returns in real terms, by taking out inflation in each year from that year's returns, and recomputing the equity risk premiums:
Note that the equity risk premiums move only slightly, because inflation finds its way into both stock and treasury returns.
Many valuation practitioners use these historical averages, when forecasting equity risk premiums in the future, but it is a practice that deserves scrutiny, partly because it is backward looking (with the expectation that things will revert back to the way they used to be), but mostly because the estimates that you get for the equity risk premium have significant error terms (see standard errors listed below the estimates in the table). Thus, if are using the average equity risk premium for the last 97 years of 5.44% (7.00%), i.e., the arithmetic or geometric averages, it behooves you to also inform users that the standard error of 2.12% will create a range of about 4% on either side of the estimate.
Pricing Questions
Coming into 2025, investors are right to be trepidatious, for many reasons, but mostly because we are coming off two extraordinarily good years for the market, and a correction seems due. That is, however, a poor basis for market timing, because stock market history is full of examples to the contrary. There are other metrics, though, which are signaling danger, and in this section, I will wrestle with what they tell us about stocks in 2025.
PE ratios and Earnings Yields
Even as we get new and updated pricing metrics, it is undeniable that the most widely used metric of stock market cheapness or expensiveness is the price earnings ratio, albeit with variations in the earning number that goes into the denominator on timing (current, last 12 months or trailing or next 12 month of forward), share count (diluted, primary) and measurement (ordinary or extraordinary). In the graph below, I focus on trailing earnings for all companies in the S&P 500 and compute the aggregated PE ratio for the index to be 24.16 at the start of 2025, higher than the average value for that ratio in every decade going back to 1970.
Just for completeness, I compute two other variants of the PE, the first using average earnings over the previous ten years (normalized) and the second using the average earnings over the last ten years, adjusted for inflation (CAPE or Shiller PE). At the start of 2025, the normalized PE and CAPE also come in at well above historical norms.
If I have terrified you with the PE story, and you have undoubtedly heard variants of this story from market experts and strategists for much of the last decade, I would hasten to add that investing on that basis would have kept you out of stocks for much of the last ten years, with catastrophic consequences for your portfolio. For some of this period, at least, you could justify the higher PE ratios with much lower treasury rates than historic norms,, and one way to see this is to compare the earnings yield, i.e., the inverse of the PE ratio, with the treasury yields, which is what I have done in the graph below:
If you compare the earnings yield to the ten-year treasury rate, you can see that for much of the last decade, going into 2022, the earnings yield, while low, was in excess of the ten-year rate. As rates have risen, though, the difference has narrowed, and at the start of 2025, the treasury rate exceeded the earnings yield. If you see market strategists or journalists talking about negative equity risk premiums, this (the difference between the earnings yield and the treasury rate) is the number that they are referencing.
At this stage, you may be ready to bail on stocks, but I have one final card to play. In a post in 2023, I talked about equity risk premiums, and the implicit assumptions that you make when you use the earning to price ratio as your measure of the expected return on stocks. It works only if you make one of two assumptions:
That there will be no growth in earnings in the future, i.e., you will earn last year's earnings every year in perpetuity, making stocks into glorified bonds.
In a more subtle variants, there will be growth, but that growth will come from investments that earn returns equal to the cost of equity.
The problem with both assumptions is that they are in conflict with the data. First, the earnings on the S&P 500 companies has increased 6.58% a year between 2000 and 2024, making the no-growth assumption a non-started. Second, the return on equity for the S&P 500 companies was 20.61% in 2023, and has averaged 16.38% since 2000, both numbers well in excess of the cost of equity.
So, what is the alternative? Starting 30 years ago, I began estimating a more complete expected return on stocks, using the S&P 500, with the level of the index standing in for the price you pay for stocks, and expected earnings and cash flows, based upon consensus estimates of earnings and cash payout ratios. I solve for an internal rate of return for stocks, based upon these expected cash flows:
The expected return from this approach will be different from the earnings to price ratio because it incorporate expected growth and changes in cash flow patterns. The critique that this approach requires assumptions about the future (growth and cash flows) is disingenuous, since the earnings yield approach makes assumptions about both as well (no growth or no excess returns), and I will wager that the full ERP approach is on more defensible ground than the earning yield approach.
Using this approach at the start of 2025 to the S&P 500, I back out an implied expect return of 8.91% for the index, and an implied equity risk premium of 4.33% (obtained by netting out the ten-year bond rate on Jan 1, 2025, of 4.58%):
You are welcome to take issue with the number that I use there, lowering the growth rates for the future or changing the assumptions about payout. That is a healthy debate, and one that provides far more room for nuance that looking at the earnings yield.
How does an implied equity risk premium play out in market level arguments? Every argument about markets (from them being in a bubble to basement level bargains) can be restated in terms of the equity risk premium. If you believe that the equity risk premium today (4.33%) is too low, you are, in effect, stating that stocks are overvalued, and if you view it as too high, you are taking the opposite position. If you are not in the market timing business, you take the current premium as a fair premium, and move on. To provide perspective on the ERP at the start of 2025, take a look at this graph, that lists implied ERP at the start of each year going back to 1960:
There is something here for almost point of view. If you are sanguine about stock market levels, you could point to the current premium (4.33%) being close to the historical average across the entire time period (4.25%). If you believe that stocks are over priced, you may base that on the current premium being lower than the average since 2005. I will not hide behind the "one hand, other hand" dance that so many strategists do. I think that we face significant volatility (inflation, tariffs, war) in the year to come, and I would be more comfortable with a higher ERP. At the same time, I don't fall into the bubble crowd, since the ERP is not 2%, as it was at the end of 1999.
Valuation Questions
Pulling together the disparate strands that are part of this post, I valued the index at the start of 2025, using the earnings expectations from analysts as the forecasted earnings for 2025 and 2026, before lowering growth rates to match the risk free rate in 2029. As the growth rates changes, I also adjust the payout ratios, given the return on equity for the S&P 500 companies:
With the assumption that the equity risk premium will climb back to 4.5%, higher than the average for the 1960-2024 period, but lower than the post-2008 average, the value that I get for the index is about 5260, about 12% lower than the index at the start of the year. Note that this is a value for the index today, and if you wanted to adopt the market strategist approach of forecasting where the index will be a year from now, you would have to grow the value at the price appreciation portion (about 7.5%) of the expected return (which is 9.08%).
As I see it, there are two major dangers that lurk, with the first being higher inflation (translating into higher treasury rates) and the second being a market crisis that will push up the equity risk premium, since with those pieces in play, the index becomes much more significantly over valued. From an earnings perspective, the risk is that future earnings will come in well below expectations, either because the economy slows or because of trade frictions. Rather than wring my hands about these uncertainties, I fell back on a tool that I use when confronted with change, which is a simulation:
Crystal Ball used for simulations
While the base case conclusion that the market is overvalued stays intact, not surprising since my distributions for the input variables were centered on my base assumptions, there is a far richer set of output. Put simply, at today's price levels, there is an 80% chance that stocks are overvalued and only a 20% chance that they are undervalued. That said, though, if you are bullish, I can see a pathway to getting to a higher value, with higher earnings, lower interest rates and a continued decline in the equity risk premium. Conversely, you are bearish, I understand your point of view, especially if you see earnings shocks (from a recession or a tariff war), rising inflation or a market crisis coming up.
I don't dish out market advice, and as one whose market timing skills are questionable, you should not take my (or anyone else's) assessments at face value, especially heading into a year, where change will be the byword. It is possible that lower taxes and less regulation may cause to come in higher than expected, and that global investment fund flows will keep interest rates and equity risk premiums low. My advice is that you download the valuation spreadsheet, change the inputs to reflect your views of the world, and value the index yourself. Good investing requires taking ownership of the decisions and judgments you make, and I am glad to provide tools that help you in that process.
For the last four decades, I have spent the first week of each year collecting and analyzing data on publicly traded companies and sharing what I find with anyone who is interested. It is the end of the first full week in 2025, and my data update for the year is now up and running, and I plan to use this post to describe my data sample, my processes for computing industry statistics and the links to finding them. I will also repeat the caveats about how and where the data is best used, that I have always added to my updates.
The Draw (and Dangers) of Data
It is the age of data, as both companies and investors claim to have tamed it to serve their commercial interests. While I believe that data can lead to better decisions, I am wary about the claims made about what it can and cannot do in terms of optimizing decision making. I find its greatest use is on two dimensions:
Fact-checking assertions: It has always been true that human beings assert beliefs as facts, but with social media at play, they can now make these assertion to much bigger audiences. In corporate finance and investing, which are areas that I work in, I find myself doing double takes as I listen to politicians, market experts and economists making statements about company and market behavior that are fairy tales, and data is often my weapon for discerning the truth.
Noise in predictions: One reason that the expert class is increasingly mistrusted is because of the unwillingness on the part of many in this class to admit to uncertainty in their forecasts for the future. Hiding behind their academic or professional credentials, they ask people to trust them to be right, but that trust has eroded. If these predictions are based upon data, as they claim they are, it is almost always the case that they come with error (noise) and that admitting to this is not a sign of weakness. In some cases, it is true that the size of that errors may be so large that those listening to the predictions may not act on them, but that is a healthy response.
As I listen to many fall under the spell of data, with AI and analytics add to its allure, I am uncomfortable with the notion that data has all of the answers, and there two reasons why:
Data can be biased: There is a widely held belief that data is objective, at least if it takes numerical form. In the hands of analysts who are biased or have agendas, data can be molded to fit pre-conceptions. I would like to claim to have no bias, but that would be a lie, since biases are often engrained and unconscious, but I have tried, as best as I can, to be transparent about the sample that I use, the data that I work with and how I compute my statistics. In some cases, that may frustrate you, if you are looking for precision, since I offer a range of values, based upon different sampling and estimation choices. Taking a look at my tax rate calculations, by industry, for US companies, int the start of 2025, I report the following tax rates across companies.
Note, that the tax rates for US companies range from 6.75% to 26.43%, depending on how I compute the rate, and which companies I use to arrive at that estimate. If you start with the pre-conception that US companies do not pay their fair share in taxes, you will latch on to the 6.75% as your estimated tax rate, whereas if you are in the camp that believes that US companies pay their fair share (or more), you may find 26.43% to be your preferred estimate.
Past versus Future: Investors and companies often base their future predictions on the past, and while that is entirely understandable, there is a reason why every investment pitch comes with the disclaimer that “past performance is not a reliable indicator of future performance”. I have written about how mean reversion is at the heart of many active investing strategies, and why assuming that history will repeat can be a mistake. Thus, as you peruse my historical data on implied equity risk premiums or PE ratios for the S&P 500 over time, you may be tempted to compute averages and use them in your investment strategies, or use my industry averages for debt ratios and pricing multiples as the target for every company in the peer group, but you should hold back.
The Sample
It is undeniable that data is more accessible and available than ever before, and I am a beneficiary. I draw my data from many raw data sources, some of which are freely available to everyone, some of which I pay for and some of which I have access to, because I work at a business school in a university. For company data, my primary source is S&P Capital IQ, augmented with data from a Bloomberg terminal. For the segment of my data that is macroeconomic, my primary source is FRED, the data set maintained by the Federal Reserve Bank, but I supplement with other data that I found online, including NAIC for bond spread data and Political Risk Services (PRS) for country risk scores.
My dataset includes all publicly traded companies listed at the start of the year, with a market price available, and there were 47810 firms in my sample, roughly in line with the sample sizes in the last few years. Not surprisingly, the company listings are across the world, and I look at the breakdown of companies, by number and market cap, by geography:
As you can see, the market cap of US companies at the start of 2025 accounted for roughly 49% of the market cap of global stocks, up from 44% at the start of 2024 and 42% at the start of 2023. In the table below, we compare the changes in regional market capitalizations (in $ millions) over time.
Breaking down companies by (S&P) sector, again both in numbers and market cap, here is what I get:
While industrials the most listed stocks, technology accounts for 21% of the market cap of all listed stocks, globally, making it the most valuable sector. Thee are wide differences across regions, though, in sector breakdown:
Much of the increase in market capitalization for US equities has come from a surging technology sector, and it is striking that Europe has the lowest percent of value from tech companies of any of the broad subgroups in this table.
I also create a more detailed breakdown of companies into 94 industry groups, loosely structured to stay with industry groupings that I originally created in the 1990s from Value Line data, to allow for comparisons across time. I know that this classification is at odds with the industry classifications based upon SIC or NAICS codes, but it works well enough for me, at least in the context of corporate finance and valuation. For some of you, my industry classifications may be overly broad, but if you want to use a more focused peer group, I am afraid that you will have to look elsewhere. The industry averages that I report are also provided using the regional breakdown above. If you want to check out which industry group a company falls into, please click on this file (a very large one that may take a while to download) for that detail.
The Variables
The variables that I report industry-average statistics for reflect my interests, and they range the spectrum, with risk, profitability, leverage, and dividend metrics thrown into the mix. Since I teach corporate finance and valuation, I find it useful to break down the data that I report based upon these groupings. The corporate finance grouping includes variables that help in the decisions that businesses need to make on investing, financing and dividends (with links to the US data for 2025, but you can find more extensive data links here.)
(If you have trouble with the links, please try a different browser)
Many of these corporate finance variables, such as the costs of equity and capital, debt ratios and accounting returns also find their way into my valuations, but I add a few variables that are more attuned to my valuation and pricing data needs as well.
(If you have trouble with the links, please try a different browser)
Not that while much of this data comes from drawn from financial statements, some of it is market-price driven (betas, standard deviations, trading data), some relates to asset classes (returns on stocks, bonds, real estate) and some are macroeconomic (interest rates, inflation and risk premiums). While some of the variables are obvious, others are subject to interpretation, and I have a glossary, where you can see the definitions that I use for the accounting variables. In addition, within each of the datasets (in excel format), you will find a page defining the variables used in that dataset.
The Timing
These datasets were all compiled in the last four days and reflect data available at the start of 2025. For market numbers, like market capitalization, interest rates and risk premiums, these numbers are current, reflecting the market's judgments at the start of 2025. For company financial numbers, I am reliant on accounting information, which gets updated on a quarterly basis. As a consequence, the accounting numbers reflect the most recent financial filings (usually September 30, 2024), and I use the trailing 12-month numbers through the most recent filing for flow numbers (income statement and cash flow statements) and the most recent balance sheet for stock numbers (balance sheet values).
While this practice may seem inconsistent, it reflects what investors in the market have available to them, to price stocks. After all, no investor has access to calendar year 2024 accounting numbers at the start of 2025, and it seems entirely consistent to me that the trailing PE ratio at the start of 2025 be computed using the price at the start of 2025 divided by the trailing income in the twelve months ending in September 2024. In the same vein, the expected growth rates for the future and earnings in forward years are obtained by looking at the most updated forecasts from analysts at the start of 2025.
Since I update the data only once a year, it will age as we go through 2025, but that aging will be most felt, if you use my pricing multiples (PE, PBV, EV to EBITDA etc.) and not so much with the accounting ratios (accounting returns). To the extent that interest rates and risk premiums will change over the course of the year, the data sets that use them (cost of capital, excess returns) allow for updating these macro numbers. In short, if the ten-year treasury rate climbs to 5% and equity risk premiums surge, you can update those numbers in the cost of capital worksheet, and get updated values.
The Estimation Process
While I compute the data variables by company, I am restricted from sharing company-specific data by my raw data providers, and most of the data I report is at the industry level. That said, I have wrestled with how best to estimate and report industry statistics, since almost every statistical measure comes with caveats. For a metric like price earnings ratios, computing an average across companies will result in sampling bias (from eliminating money-losing firms) and be skewed by outliers in one direction (mostly positive, since PE ratios cannot be negative). Since this problem occurs across almost all the variables, I use an aggregated variant, where with PE, for instance, I aggregate the market capitalization of all the companies (including money losing firms) in an industry grouping and divide by the aggregated net income of all the companies, including money losers.
Since I include all publicly traded firms in my sample, with disclosure requirements varying across firms, there are variables where the data is missing or not disclosed. Rather than throw out these firms from the sample entirely, I keep them in my universe, but report values for only the firms with non-missing data. One example is my data on employees, a dataset that I added two years ago, where I report statistics like revenue per employee and compensation statistics. Since this is not a data item that is disclosed voluntarily only by some firms, the statistics are less reliable than on where there is universal disclosure.
On an upbeat note, and speaking from the perspective of someone who has been doing this for a few decades, accounting standards around the world are less divergent now than in the past, and the data, even in small emerging markets, has far fewer missing items than ten or twenty years ago.
Accessing and Using the Data
The data that you will find on my website is for public consumption, and I have tried to organize it to make it easily accessible on my webpage. Note that the current year’s data can be accessed here:
If you click on a link and it does not work, please try a different browser, since Google Chrome, in particular, has had issues with downloads on my server.
If you are interested in getting the data from previous years, it should be available in the archived data section on my webpage:
This data goes back more than twenty years, for some data items and for US data, but only a decade or so for global markets.
Finally, the data is intended primarily for practitioners in corporate finance and valuation, and I hope that I can save you some time and help in valuations in real time. It is worth emphasizing that every data item on my page comes from public sources, and that anyone with time and access to data can recreate it. For a complete reading of data usage, try this link:
If you are in a regulatory or legal dispute, and you are using my data to make your case, you are welcome to do so, but please do not drag me into the fight. As for acknowledgements when using the data, I will repeat that I said in prior years. If you use my data and want to acknowledge that usage, I thank you, but if you skip that acknowledgement, I will not view it as a slight, and I certainly am not going to threaten you with legal consequences.
As a final note, please recognize that this I don't have a team working for me, and while that gives me the benefit of controlling the process, unlike the pope, I am extremely fallible. If you find mistakes or missing links, please let me know and I will fix them as quickly as I can. Finally, I have no desire to become a data service, and I cannot meet requests for customized data, no matter how reasonable they may be. I am sorry!
