So, what works in forecasting? Lets review.

1. Low Price to Asset Value

The principle here is that if you buy a stock trading at less than its book value (the cost it paid for its own assets) then it will eventually reach at least that value. Or stocks trading at discount to net current assets (ie cash and other assets which can be turned to cash in a year, less liabilities).

2. Low Price to Earnings

Also included in this category are high dividend yield stocks and low prices in relation to cash flow (earnings plus depreciation expenses). A popular one with Benjamin Graham and Warren Buffett but adding that they want to see growth too.

3. Directors buying own stocks

The theory is that ‘insight information’ these individuals have should afford a clue as to their quality and likely price rising.

4. A significant decline in a stock’s price

Do companies whose share price has fallen, do so because earnings have dropped and if so, do earnings revert to mean, and so therefore then see a share price rise?

5. Small market capitalization

Allowing for risk, do smaller companies outperform?

Low Price to Asset Value

Net current asset value approach favoured by Benjamin Graham calls for the purchase of stocks which are priced at 66% or less of a company’s underlying current assets.

Graham’s own asset management firm produced results of 20% annually using this technique. Essentially if a company’s current asset are $100 per share and the sum of current liabilities, long-term debt, stock and other liabilities is $40, then Graham would not pay more than 66% of $60, or $40, for the stock. This method appears to work based on other studies too. Indeed, it even ignores real estate values too.

Such companies often have their own officers buying stock, low p/e. low price/sales. They sometimes have had a fall in their share price too – that answers the question of when is a fall in a share price good – if it means a low price to asset value.

Low Price to Book Value

Research shows stocks with a low price to book value ratio outperform others. Below is from firm Tweedy, Browne.

Market Capitalisation

What if as well as price to book value, or net current assets, we also add market capitalization – does that improve our forecasting abilities? Fama and French found the following:

So we certainly in our investment criteria want to account for market capitalization.

So are the returns due to greater risk? Are they consistent? The tables below show low price to book is a consistent and better risk-adjusted method of forecasting than high price to book.

Low Price to Earnings

How good is this as a forecaster? Let us not just take magazine and TV talk for it, but take sound academic proof and research on it. Below is a result from Ibbotson in a working paper for Yale School of Management.

Other findings are that companies selling at low price/earnings ratios often have above average dividend yields. They also retain a part of their earnings to reinvest in the business. A low p/e by definition must provide an above average dividend yield or an above average retained earnings yield, or both.

So we know some of the old methods for forecasting work – and now we know why. But what about forecasting based on p/e and market cap. Again, see the table below to answer that.

The bottom line is that $1m invested in the lowest price/earnings ratio companies with the lowest market cap in 1969 would have increased to $29.8m by 1989. More updated information proves this too:

Insider Stocks

How good are selection criteria based on ‘insiders’ buying their own company stock? Various studies, show below, suggest this is a good way to pick stocks.

Stocks which have declined in price

Do poor performing stocks end up outperforming? Its counter-intuitive, yet we can all think of great examples.

The research below suggests they do well. The findings are from Power, Lonie from Dundee University, ‘The Over-reaction Effect – Some UK evidence’.

Forecasting consistency: Patel’s Sixth Principle of Forecasting: Consistency of returns are worth more to investors than the return itself, and they should be to the investor too.

I manage other people’s money. I know my investors may ask me about the performance last year, and the year before, and we may think they care about it, but they really care about consistency. I know this, because history is soon forgotten when you send them a monthly update. They weigh consistency more than academic literature suggests.

The problem with any forecasting method or screen based on criteria that have worked, is not that what has worked in the past may not work in the future, but that the definition of ‘worked’ could mean negative returns for two consecutive years. Trust me when it comes to the forecasting business you are not a good forecaster if you promise to deliver 20 per annum over a 5 year period, if in your first year you are down 20%. You may well make over 20% annually thereafter. But two forecasting problems, real world problems, not academic theoretical problems will face you:

  1. your own trust in your own method will be at breaking point
  2. your investors if you have any certainly will not think you a good forecaster.

Remember Buffett is astounding not because he can return 30% on average each year – he can’t – many hedge fund managers can, but they are not known or the second richest men on the planet. He is well regarded because he can be consistent with under a 25% return annually.

(By Alpesh Patel. Analysis here appeared originally in the Financial Times 'Diary of an Internet Trader column by the author and is used for his investment software . You can learn more also at