How To Identify & Use Good Historical Data For A Financial Model

Financial analysts often struggle to find a sound basis for forecasting future values for financial and operating results in a financial model. We have looked at the common qualitative and quantitative methods for forecasting before, and it will be useful for financial analysts to better understand the importance of having good quality historical data when applying the quantitative or statistical forecasting based techniques.

Historical data sets of the business or industry are most often used as the “quantifiable underlying base” to statistically forecast key historical variables into the future.

As with all statistical forecasting techniques, the basic premise is to analyze this historical data to identify a mathematical relationship between the variable we wish to forecast (i.e. the dependent variable) and the variables that are being used to calculate the forecasted result (i.e. the independent variables). We can then use this mathematical relationship to forecast the dependent variable in the future.

There are 2 questions that a financial analyst should consider before applying these statistical forecasting techniques to forecast the future values of the dependant variable for a financial model:

Do we have enough data to develop a forecast from?
Is a statistically relevant historical relationship likely to hold in the future?

Lets now examine how a financial analyst can proceed to address these 2 questions.

Do we have enough data to develop a forecast from?

Or in other words, whether there is sufficient historical data to determine the relationship with a reasonable degree of accuracy.

To use an example, in many financial models, particularly financial models built around annual time periods, it is often very difficult to obtain reliable historical data going back by more than 5 or 6 years. A shortened historical data set will basically mean that the financial analyst will not feel confident of his / her ability to forecast forward over any substantial length of time.

This is due to the statistical fact that the further out you forecast in a financial model, the larger the forecast error. As a reasonableness test, a financial analyst could apply the statistical test of determining a confidence level for a forecast using the t statistic and the degrees of freedom. An introductory learning program on business statistics will provide a more detailed description of this technique.

Is a statistically relevant historical relationship likely to hold in the future?

Even if you do obtain a significant statistical relationship from the historical data, there is still the basic issue of whether a historical relationship between a dependent variable and independent variables will continue to hold into the future.

Therefore, in answering this question, the financial analyst would look to whether there have been any material changes in the historical data. If there has been a significant structural change then there may be no relationship between what has happened historically and what is likely to happen in the future.

Using another example, assume a financial analyst is trying to build a financial model to forecast the sales of an expensive piece of computer hardware equipment over time, using historical data going back 20 years.

The financial analyst came across in his industry research that 9 years ago, the monopoly of large manufacturers which controlled the sale prices of the computer equipment was dissolved due to regulatory pressure. This invited a host of low-cost manufacturers who started selling the product at various competing price points, thus changing the basic structure of the industry.

In this case, it is likely that any historical data on sales price point prior to the collapse of the cartel 9 years ago is irrelevant to the future forecast of the product.

In conclusion…

Assuming we’ve assessed and verified the accuracy, credibility and sufficiency of historical data, we can then proceed to determine if there is a statistically significant historical relationship between the dependent and independent variables in the data and apply statistical techniques for financial model forecasting.

If you're new to Financial Modeling Guide, you may want to join our sister site, the Finance 3.0 network, to enroll in online financial training programs, ask questions, attend seminars, connect with others and download free finance spreadsheet templates about financial modeling, financial management, corporate finance & valuation, quantitative finance and accounting.
Over 65,000 finance professionals, entrepreneurs, analysts and investors from 195 countries have benefited.
Don't miss out on this outstanding opportunity for high quality, globally recognized financial education: Sign Up For Free Now!