Forecasting Finances

In this lesson, you’re expected to learn about:
– analyzing data effectively
– interpreting the past
– predicting the future without preconceptions
Analyzing Data
In this lesson, we’ll cover using current and historical data about a company, the industry and the economy to predict the future.

Of course, your predictions may be wrong, but you can also use the data to determine the probability of being wrong and by how much you may be off the mark. As a result, we use a lot of basic statistics and probability, much of which we saw in the Statistics & Big Data module.

To forecast finances successfully, you have to let go of your preconceived notions about what you think the future will look like and instead allow the information you have available to guide your ideas. Too many people working in forecasting and projections let their established ideas and beliefs get in the way of what may otherwise be very promising data.

To start, disregard what you think you know about the company you’re studying and discover everything afresh, starting with the company’s historical data. After you understand the data, you can then use everything you already know about the company to determine the reason it’s performing in that manner and only then can you predict what’s going to happen.

Confirmation Bias

One common mistake, known as confirmation bias, is that people tend to look at data simply to confirm what they think they already know.

Confirmation bias is the normal result of the human tendency to rely on patterns and trends to make sense of the world, but it can also cause you to be fooled relatively easily. So, analyze the data first and then worry about using what you know to explain that data.

Collecting Data

Before you can analyze any of the data that’s going to help you project your company’s future financial performance, you need to collect that data. Thankfully, with the help of the internet, this task is fairly easy.

You can also collect the required data by requesting it directly from the company concerned or going to a financial adviser to get the information you need.

Data collection can be a pretty tedious and sometimes long, drawn-out process. Narrowing down the exact type of information you’re looking for helps a lot and results in people generally being more forthcoming with the information.

The great thing about financial data is that everything about the data itself is the same. It’s all measured in money, and so you don’t have to worry about the technical details of the different types of data.
Knowing where to look

The vast majority of financial data you want about any business comes from a few primary locations:

1) Financial reports: Financial reports are usually the first place to look because they’re easy to find and already formatted in a way that’s relatively simple to analyze. Not just the annual reports — quarterly and monthly reports are important as well. Sometimes you can just go to the relevant company’s website and download all the financial information directly. Published reports are a reliable source of information.

2) Reports regarding stock, production and employment: Companies, particularly larger ones, occasionally distribute reports on stock, production and employment, especially when prompted by organizations that are attempting to compile economic reports.

3) Accounting records: If you’re able to acquire the company’s accounting books and records, these sources are easily the most comprehensive and detailed data you can find. Though, getting your hands on such information is often difficult.

4) Internet sources: For information about share prices (critical for many financial calculations), lots of websites are available, all providing basically the same information.

Comparing your data

As with any financial data, you probably benefit by collecting the same information from several other companies in the same industry for comparison.

You should also research data on the national economy. There are a variety of sources for this such as:
– International Monetary Fund (IMF)
– World Bank

Finding an Average  

After you gather all your information, you need to work out what to do with it. You have to do some simple descriptive calculations of statistics and probability.

In other words, you measure the likelihood of an event occurring using information about performance and relationships between variables (i.e., information that is subject to change).

When you have a lot of different values for a variable, finding an average tells you the middle value — in other words, what’s typical. Averages fall into different types, each with its own strengths and weaknesses, but in financial equations the vast majority of averages are the mean average.

Weighted Average

To look at a weighted average (an average that takes into account differences in the importance of each value), you attach a weight to each value.

For instance, if one of the values in a set of data is worth 60% of the entire sample and the rest weighted equally at 10% each, the average changes a bit:
= 1 (0.1) + 2 (0.1) + 3 (0.1) + 4 (0.1) + 5 (0.6)
= 0.1 + 0.2 + 0.3 + 0.4 + 3.0 = 4

The weighted average is 4 because the value 5 has more weight than the other values, bringing the average up compared to the standard mean average (3).

Most people use the weighted average in situations where an investment portfolio has different proportions of investments or when they’re accounting for time-weighted averages, where more recent values are more important than historical ones.
Moving Averages

A more common method used in financial analysis and projections is moving averages, which takes the average from a predetermined number of days prior to a given day.

For example, for a three-day moving average on Wednesday, you include data going back to Monday; for Thursday, you collect all the data going as far back as Tuesday; and for Friday, you go back to Wednesday. This data helps illustrate whether the mean is increasing or decreasing over time.

Measuring Data Distribution

Obviously, not all the numbers in a data set are going to be exactly the same as the average. You can measure the manner in which data is distributed around the average in different ways.

Say that the average net profit of a company is $10,000. That’s great, but it doesn’t tell you whether that number changes much: the company may consistently earn $10,000 every year, or it may earn $0 in the year before and $20,000 the year after.

This information is the sort of thing worth knowing, and we describe two ways in which you can measure it.

1) Range

Range is simply the difference between the largest and smallest values. So, if a company has profits of $10,000 and $20,000, you can say that it has a two-year range of $10,000, or 100 per cent.

If you’re looking at the range for the company’s profits over the last 20 years, you may want to pay attention to its interquartile range (which simply means the range of the middle 50% of values).

This approach allows you to make sure that the company didn’t experience unusually high or low profits in certain years, which would affect the data.

Interquartile Range

To find the interquartile range, you take the profits from all the years and put them in numerical order, divide them into four equal pieces and then take the range of the middle two pieces.

So, if a company’s profits have a range of $100,000 but an interquartile range of only $20,000, you may think that the company had extreme variation in its profits in some of those years.

These ranges are often illustrated on graphs in a couple ways:
– Box plots
– Bollinger bands

2) Standard Deviation

Standard deviation is another measure of distribution, represented by σ (sigma). It’s a concept used quite frequently in equations, and here’s how you calculate it:

1. Calculate the mean average.
2. Subtract each value from the mean.
3. Square each difference.
4. Add the squares together.
5. Divide the answer by the number of values.
6. Take the square root of the answer from Step 5.

What standard deviation tells you is the extent to which results are spread out from the average — for example, if the temperature in one month stays about the same each day, the standard deviation will be low.

Understanding Probability

Probability theory is pretty easy. The total probabilities of an event occurring or not always equal 100 per cent. If you have a 10 per cent probability that something may happen, you have a 90 per cent probability that it won’t.

The simplest example is a coin toss. You have a 50 per cent probability that the coin is going to land on either side because only two options exist. Each time you flip that coin, you have a 50 per cent probability of it being heads or tails.

Normal Distribution

When you apply probability theory to the standard deviation, you end up with something called a normal distribution.

The normal distribution, shown in the figure below, has a lot of very important traits, but all you really need to know is the relationship between standard deviation, probability and the distribution of data.

The percentages in the curve tell you what percentages of the data are included within the number of standard deviation units listed at the bottom. After you calculate the standard deviation and the mean, you can work out probability easily.

Importance of Normal Distribution

For example, say that you have a mean of 4 and a standard deviation of 1. According to the graph, 34 per cent of all values will be between 5 and 6, 68 per cent of all cases will be between 4 and 6 and so on.

You care about normal distribution because probability calculations are used frequently in financial forecasts. Imagine that you want to predict the most probable percentage drop in the stock market as a result of an increase in interest rates. By collecting historical data and determining the mean and standard deviations, you can estimate the likely range to any percentage of probability you like.

The more certain you want to be, the wider your range is going to be, because you have to account for a greater range of data that encompasses a particular level of probability.

Bayesian Probability

You can take the normal distribution calculation in the preceding section a step further. Imagine that you want to know the probability that, given the event that the stock market drops by 1-2 %, a specific company’s share price is also going to drop by 1-2%? You can come up with the answer to this question by using Bayesian Probability.

This equation says that in order to calculate the probability of event A happening conditionally of event B, you carry out the following steps:

1. Take the probability of event B happening as a result of event A and multiply that amount by the probability of event A.
2. Divide the answer by the probability of event B happening.

Need for Historical Data

Most people are obsessed with their money and spend a lot of time and resources tracking and recording data. As a result, just about all the historical data you can ever want regarding corporate finance is already collected and compiled. You don’t have to do any of that time-consuming research; you just need to collect the data that others have found.

When reviewing historical data you need to shut off everything you think you know. When dealing with issues of uncertainty, such as forecasting, behavioral mistakes tend to be amplified.* So go into your research with an open mind, and always be on the lookout for something interesting that others may not have noticed in order to give yourself a financial edge.

* Next week, we’ll discuss the concept of Behavioral Finance.
Finding Trends and Patterns

When reviewing historical data, your first job is to look for trends and patterns. If you can identify trends that are occurring and any cyclical patterns that have happened in the past, you gain important insight into what will (or at least may) happen in the future.

Start with patterns, for example. You can usually best explore patterns by plotting your data on a graph. Try several different graphs and really look at each of them to see whether you can recognize any patterns that begin to emerge.

Not all patterns are obvious or simple, but the basic premise is the same: you’re looking for any patterns that allow you to predict what’s going to happen in the future of your company’s finances.


The goal of using regression is to look at historical information to determine whether any variables are influencing financial movements.

Today this process typically uses highly advanced computer programs, such as analytics software and databases, to perform something called data mining. Basically, data mining works by including all the data you can possibly get your hands on and letting a computer program decide whether any correlation exists between what you’re trying to forecast and other variables.


For example, you may find that your company’s costs increase with the temperature outside. As the temperature increases, so do total costs; as temperature decreases, the company’s costs also decrease. You may even find that, on average, costs change by 1% for every 3% change in temperature. This relationship is called a correlation.

Note that a correlation doesn’t mean that the temperature is causing a price increase — just that the two are related. Correlation does not imply causation.

If temperature and cost are correlated, the relationship may look something like the figure below. The points on the graph are the actual values – you plot them as you would on any graph.

The regression line going through them illustrates the proportion of the relationship. (In this case, a one-third slope indicates that for every unit increase in cost, temperature increases by 3 units.)

Consider the following attributes:
– Positive correlation: As one factor increases, the other increases as well.
– Negative correlation: One factor decreases as the other increases.

Knowing what to do with correlations

Ideally, if you can find a relationship, you want to be able to use it to make financial predictions.

Multivariate Regression
You can also use multiple variables to create more accurate correlations. These multivariate regressions attempt to show how each variable has an influence on what you’re measuring.

When used together, you can create an even more accurate model that not only explains what’s causing changes in the thing you’re measuring, but also how much of a role each variable plays and how you can use that to predict what will happen in the future.

With regards to investing, any correlations that allow you to predict movements in the share price are highly prized.

Predicting the future: forecasting
Financial forecasts are used in just about every aspect of corporate finance. Budgeting, investing, risk assessment, financing, stock management, production schedules etc.

People obsess over money, and they want to know everything about it, including what will happen in the future. After you analyze your data, you can provide them with predictions of the future.

The forecast itself is often nothing more than a prediction of what’s going to happen, typically including the probability of the prediction being correct and a range of other values that may also occur, with an explanation for the deviation.

Forecasters prefer to try and provide more information than just the basics. They include any information that may be potentially useful for making decisions or back-up plans.

Forecasting finances is a bit like forecasting the weather; you like to know if the probability of rain is low, but unless it’s 0% probability, you should probably make back-up arrangements as well.

Using Statistics and Probability 

Simply put, to forecast your finances you look out for trends, patterns and relationships, determine the probability of these factors influencing a particular outcome and use that to model your forecast.

For instance, if government indicators predict that the economy is going to grow by 4% next year and you’ve assessed a correlative relationship of economic and sales growth predicted by indices at half of the government indicators, you should predict that the economic growth will contribute to a 2% sales increase next year.

Does that mean that sales will increase 2% next year? Only if nothing else influences your sales at all, because other factors may make sales higher or lower, but the economic growth will have a bit of a positive influence on your sales.

Based on consistent trends over each month of the last three years of a steady 1% monthly sales increase, you may predict that you’ll continue to see steady growth over the next several years, but with a 68% probability of slowed growth because you find patterns where sales slowed down every fourth year.

Even if you can’t work out what variables influenced that slowed growth, after calculating the probability of it you can determine that your sales have a definite possibility of a temporary slow-down.

Predicting Movements

In the stock market, the two things that are most commonly used to predict movements are earnings and price.

However, using these two items as predictors is not ideal, because both tend to be too volatile and too easily manipulated to be useful indicators.

So what are good indicators?
A somewhat more accurate indicator is the yield on Treasury bonds. This yield tends to increase and decrease in a generally similar way to national gross domestic product, but two to four years earlier. Ratios such as price-to-earnings are also quite popular for predicting stock-market movements.

Seeking a Precedent: Reference Class Forecasting

Reference class forecasting involves finding a similar precedent set in the past for what you’re trying to predict and then using the outcome of that scenario to check whether your forecast is reasonable compared to what happened historically.

Because reference class forecasting is very prone to variations, given that not each situation is exactly the same, performing the forecast first helps you avoid bias or guiding scenarios (where your opinion is shaped by preconceived notions rather than the data itself).

When you do the reference class forecast and the data doesn’t match expectations based on your reference, you can determine why it’s different and alter your forecast as necessary.

Evaluating Forecast Performance

You can use two primary methods to evaluate financial forecasting performance: time and accuracy.

A forecaster is considered more successful when he’s able to predict very closely when something will occur or very closely the degree to which something is going to change. If a forecaster predicts that revenues will jump in July, but sales drop in July only to jump in August, the forecaster isn’t very accurate.

Of course, a few variables (such as production capacity, for example) influence how those differences in forecasts should be interpreted. If sales jump by 11% instead of 10% and the company isn’t ready to handle that extra 1% jump, the forecaster will be held responsible by management.

[Optional] Liquidity Ratios
Check out this article to learn more:
Jim Rohn Sứ mệnh khởi nghiệp