Predicting stock market returns

At a glance

  • While predicting market performance is nearly impossible, there are lessons to be learnt from looking at past performance
  • We take the predictive power of three popular valuation metrics; Price-to-Earnings, Price-to-Book and Dividend Yield
  • P/E as a valuation metric performs better than P/B and yield as a lead indicator of annual returns
  • As of early Jan 2015, the Nifty is trading at 21.2x earnings, more expensive than 80% of all trading days in the last 16 years
  • Higher the P/E at the time of purchase, lower the returns achieved for the Nifty
  • Current P/E levels indicate marginally negative returns for the next year

The four most dangerous words in investing are: this time it’s different – John Templeton

History suggests, those who assume that good times or bad, will continue to persist indefinitely, typically “get schooled” by Mr. Market. Predictions about market levels and stock prices, like Niels Bohr said, are very difficult, especially about the future. But neglecting to atleast try to understand, what transpired in the past, is like packing for a vacation destination, while ignoring the weather report.

My last post of 2014, summarized the buoyant year for Indian markets in three charts. In another earlier post: ‘How to prepare for a market correction‘, I wrote how markets being up 40% YTD till November 2014, made the calm investor slightly uneasy. To an extent, that is the lot of the value investor, who’s most at ease during widespread doubts about the viability of equities as an investment avenue.

So, while investors (and companies) wait for the promised slew of reforms and return of confidence to boost corporate performance,  should we keep buying Indian equities, taking the turnaround for granted, or risk missing out on the big bull run by being cautious and sitting it out for a while?

Some analysis of historic data might help. I tracked returns offered by the Nifty and CNX 500 over 15 years superimposed on how expensive or cheap the market was as defined by three common valuation metrics: Price-to-Book, Dividend Yield and Price-to-Earnings.

Finding #1: Nifty more expensive now than 80% of the time

In a previous post, I mentioned why it doesn’t make sense to look at absolute index values (since they are bound to go up as underlying companies grow). A normalized valuation metric like Price-to-Earnings provides a view on how (in)expensive the market is compared to historical trends. Chart above shows that as of Jan 2015, the NIFTY is trading at 21.2x earnings, making it more expensive than 80% of the time in the last 16 years.

Finding #2: Higher the P/E at purchase, lower the return

The Calm Investor

Since P/E represents the number of years of current earnings you are prepared to pay for a stock, a higher P/E represents higher expectations and optimism about where the earnings are likely to go. However, historically, the more optimistic a purchase, the higher the disappointment. Therefore, buying the index when trading at 13x earnings on average has given 60%+ returns in a year, while paying more than 23x earnings has typically resulted in erosion of 25% of your capital.

Finding #3: Price-to-Earnings a better predictor of returns

Running a regression on the 1 year returns against current values of P/E, P/B and Dividend Yield tells us the strength of the relationship between these three measures and the returns. The closer the R-square is to 1, the stronger the relationship. (Here’s a quick primer on R-squared)

  1. At the index level Price-to-Book (Ratio of stock price to Book value of assets) has a weak link with expected returns and should be ignored
  2. Both Dividend Yield (dividends earned as a percentage of the stock price) and Price-to-Earnings (Ratio of stock price to Earnings per share) show decent (note that they are nowhere near perfect) ability to predict the returns 1 year into the future

Finding #4: Current earnings unlikely to sustain returns

Charts compares actual 1 year return against predicted return based on Price-to-Earnings (P/E). Meaning plugging the current P/E predicts what the return will be 1 year in the future

  1. The predictor is far from perfect (no predictor can possibly be), but the red line broadly tracks the blue for both the Nifty and the CNX 500 indices over a 16 year period which is commendable
  2. Currently, both the benchmark NIFTY and the broader-based CNX500 project negative returns for the next year (Jan 2015-16), -1% and -10% respectively.

Therefore, for the calm investor, it will be vital to identify those stocks that currently trade at reasonable prices for quality and sustainable earnings.

11 thoughts on “Predicting stock market returns

  • January 13, 2015 at 3:47 pm

    Trying to predict one year return does not work, while i agree that valuation based metrics will broadly be right ( P/e or P/B etc) the process lasts for long, Would be better if you see the 3/5 year returns post investing in high pe markets.

  • January 13, 2015 at 9:18 pm

    Fair point Shankar. Thanks for your comment. Longer terms results might show an even stronger relationship with valuation. However we’re constrained by availability of data. Unlike the U.S. where there’s nearly a century of data available on these parameters.

  • January 14, 2015 at 8:32 pm

    Be careful – you are using the same data to predict as the model itself.

    You used past data to determine how stocks have behaved 1 year in the future based on the P/E that the index had (average ranges).

    Then you plot the “prediction” of that past data as a backcasted chart. Obviously it will show the same curve, because you used the very same data to determine the ranges in the first place 🙂

  • January 15, 2015 at 10:03 am

    Thanks for bringing it up Deepak! Appreciate your feedback. I used this mechanism for simplicity of explanation.
    This is how I originally did it:
    Broke up historic data into test (1999 – 2010) v/s control (2011 onwards). The test data revealed R-square of 0.53. Values (intercept and slope) emerging from the test data fit the control (since 2011) returns fairly well (correlation of 0.78). Hence the assertion that the predicted returns for 2015 are marginally and substantially negative for NIFTY and CNX 500 respectively.

  • Pingback: Predicting market returns? Not so fast! | The Calm Investor

  • Pingback: Three value investing lessons from Graham and Dodd | The Calm Investor

  • April 8, 2015 at 2:24 pm

    Could you explain, how have you predicted the Nifty with the P/E? At what time do you create this prediction, I mean how much in advance? I have read many articles on P/E relation with future return but no one is so close as your prediction?

  • April 9, 2015 at 7:19 pm

    Hi Girish, I took 15 years of daily data on P/E available for NIFTY. Using the NIFTY level for 1 year from each date, calculated the 1 year return. Then ran a regression with return as the dependent variable and the P/E as independent (meaning found a relationship between current P/E and 1 year return). The relationship has R-squared of about 50% which means the current P/E explains about 50% of the future return, far from perfect but is a useful input. I did another post to explain this a bit more (

    Let me know if you’d like to discuss further.

  • Pingback: Why the Nifty is falling and the futility of forecasts | The Calm Investor

  • Pingback: The NIFTY in 2016 | The Calm Investor

  • Pingback: 12 Axioms of Risk Reward followed by generations of Swiss Bankers | The Calm Investor

What do you think?