How forecasters can make sharper profit predictions

A track record yields better data when broken into components

Profit forecasting is something of a black art practised by financial analysts.

When Wall Street pundits come out with earnings forecasts for publicly listed companies, there is little transparency in the methodology and calculations they have used to arrive at the numbers.

Because the analysts are competing with each other for recognition, most often any intellectual property they use in their forecasts is undisclosed, with the result that the numbers appear to come out of a 'black box'.

What is probable, however, is that any analyst using rigour in their research will look at the historical context around the stock they are reporting on. This means looking not just at the company itself, but the market it operates in and the industry it is a part of.

Understanding how a company has performed historically against its peers is one of many first steps in assessing its performance, and arriving at a view on its future earnings. This is one of the first lessons taught to financial analysts.

But while analysts do look at these factors there hasn't been any hard or proven accounting methodology they can use to formalise this analysis to arrive at their forecasts.

Andrew Jackson, a senior lecturer in the school of accounting at UNSW Business School, has sought to address this issue in a research collaboration with Marlene Plumlee from the University of Utah, and Brian Rountree from Rice University.

Their paper, Decomposing the market, industry and firm components of profitability: Implications for forecasts of profitability, creates a methodology that breaks down a firm's profitability into market, industry and firm specific (or idiosyncratic) components, and then uses these breakdowns in profit forecasting.  

"Our approach is based on an academic paper published back in 1967, but since then we are not aware of any research that has attempted to specifically assess each of these different components," says Jackson.

"Historical analysis has long been part of estimating future earnings and company value, but where our work takes this further is in arriving at beta measures for the market, industry and firm components and then using these to construct forecasts."

'If you are using our methodology you are going to get a more accurate forecast around 68.6% of the time'

ANDREW JACKSON

Decomposing components

The researchers began by looking at 244,492 quarterly results posted by 9323 publicly listed US companies, and then narrowed this down to a sample of 3045 firms and their 75,652 quarterly results dated from 1976 through to 2014.

The first step was to use this data to estimate betas that captured the historical relationship between the return on net operating assets (RNOA) for the particular company and the RNOA for the US market and the specific industry the company was operating in.

"We take the firm's historical correlation with the market and from that we are able to get a marketing earnings beta," says Jackson.

"We strip away that component from the firm's profitability, and then we look at the remaining component and its historical sensitivity to industry sector earnings.

"From that we get an industry earnings beta and from that we can understand what component of that is attributable to industry-wide earnings.

"And the remainder is the idiosyncratic firm component, the firm-specific beta which is attributable to specific factors such as leadership, or the development of particular intellectual property by that company."

To calculate these betas the researchers used 20 quarters of historical data for each company, or five years' worth of results, but in doing the forecasting they used another 10 years of data to calculate the forecasting parameters, so 15 years of data in total.

The ultimate aim of the research is to understand if, by 'decomposing' RNOA into the three components, improvements in forecasting changes in future profitability can be gained and the evidence suggests that it can, though the methodology is more accurate when applied to profit-making companies than those making a loss.

"If you are using our methodology you are going to get a more accurate forecast around 68.6% of the time," says Jackson.

"Anything above 50% is considered to be useful, so this is very encouraging. However, you are only going to get a more accurate forecast 36% of the time if you apply the methodology to loss-making firms."

'This definitely has potential for Wall Street analysts and equities analysts in other markets' –

ANDREW JACKSON

Months rather than years

At Australian fund manager IFM Investors in Sydney, Jim Copland is one of the team responsible for the analysis of 'small cap' firms listed on the Australian Securities Exchange.

Copland agrees that profit forecasting can sometimes be seen as a black box, but says that at IFM the analysts use an in-house template as their starting point when they financially model any company.

This comprises data from financial statements, cash flow and the company's balance sheet, but the 'front end' of the model is different for each industry sector and is one of the key filters for analysis.

"Our framework for looking at stocks is very industry-specific, and we also take a macro view which factors in the economy and factors such as interest rates, currency exposure and inflation," Copland says.

"All of these things go into the mix, along with a view on the industry sector and expectations of margins, and then we seek to understand the management of the company and how they execute."

If there is a secret, says Copland, it is to keep the analysis simple enough so there are up to four 'levers' which are identified as drivers of performance, and ultimately are used in forecasts.

He agrees that historical data is useful, and sees the validity in Jackson and his team using 15 years of data in their methodology.

"In my universe, however, a lot of the companies haven't been around that long," says Copland.

"We are often looking at companies which have been around for months rather than years, so there is less historical data and you have to come up with your own assumptions when you attempt forward looking forecasts."

Commonly used tool

For companies with extended histories, Jackson believes his methodology has the potential to take forecasting out of the black box and become a commonly used tool.

He has written code within a software package for its application, and anyone reading the research paper who had the skills to do so could write code which does the same thing.

"This definitely has potential for Wall Street analysts and equities analysts in other markets," Jackson says.

"If they are not already taking advantage of this approach then there is scope to use our methodology in order to deliver more accurate forecasting."

Beyond profit forecasting, Jackson and his colleagues are continuing their research in the belief that it can be applied to broader accounting issues.

"Our research in these areas is at an early point, but we see the potential to explore this even further and create something which can have a use beyond profit forecasting," he says.