Comprehensive Reviewer: Financial Forecasting - The Quantitative Approach
Goal: After reviewing this material, you should be able to:
* Explain the Quantitative Approach of Forecasting.
* Apply various techniques of Quantitative Forecasting.
What is Financial Forecasting?
Imagine trying to guess what your favorite sports team's score will be in their next game, or
how much rain will fall next month. That's forecasting – trying to predict what will happen in
the future.
Financial forecasting is specifically about predicting future financial outcomes for a business.
This could be future sales, costs, profits, or even how much cash the business will have.
Two Main Ways to Forecast:
* Qualitative Approach: This relies on opinions, expert judgments, and non-numerical
information. Think of it as asking experienced people what they think will happen. (e.g.,
Sales force polling, expert opinion, Delphi technique).
* Quantitative Approach: This is what we'll focus on! It uses numbers, historical data, and
mathematical models to make predictions. It's like looking at past weather patterns to predict
tomorrow's rain.
The Quantitative Approach of Forecasting: Simple Explanation
The quantitative approach to financial forecasting is all about using past numerical data to
predict future numbers. It assumes that patterns and relationships that existed in the past will
continue, at least to some extent, into the future.
Why use it?
* It's objective: Less prone to personal biases because it's based on calculations.
* It's systematic: You follow clear steps and formulas.
* It's useful for stable situations: Works best when past trends are likely to continue.
When is it useful?
* When you have a lot of historical data available.
* When past patterns are expected to repeat.
* For short-to-medium term forecasts (though some techniques can be used for longer
terms).
Key Techniques of Quantitative Forecasting
1. Time Series Analysis
This is the most common type of quantitative forecasting. It involves looking at data points
collected over a period of time (like monthly sales figures for the past five years) to identify
patterns and trends.
Think of it like looking at a graph of your company's sales over the last few years. Do they go
up? Down? Are there ups and downs every year at the same time?
Common Time Series Patterns:
* Trend: A long-term increase or decrease in the data (e.g., sales consistently growing each
year).
* Seasonality: Patterns that repeat over a fixed period (e.g., higher sales during holidays,
lower sales in summer).
* Cyclical: Wavelike patterns that repeat, but not necessarily over fixed periods (often tied to
economic cycles).
* Random/Irregular: Unpredictable fluctuations caused by random events (e.g., a sudden
natural disaster affecting sales).
a. Naive Method
Concept: The simplest forecasting method! It assumes that the next period's forecast will be
exactly the same as the most recent actual value.
In Simple Words: "Whatever happened last time, will happen again this time."
When to Use:
* When there's very little data.
* When there are no clear trends or seasonal patterns.
* As a baseline to compare more complex forecasts against.
Formula:
Forecast for next period = Actual value of the most recent period
Example/Application:
Let's say a small store's actual sales for May were ₱50,000.
Using the Naive Method, the forecast for June's sales would be ₱50,000.
If June's actual sales turn out to be ₱52,000, then the forecast for July would be ₱52,000.
b. Simple Moving Average (SMA)
Concept: This method forecasts the next period's value by averaging the actual values from
a specific number of previous periods. It smooths out random ups and downs.
In Simple Words: "Let's average the last few results to guess the next one."
When to Use:
* When demand is relatively stable (no strong trend or seasonality).
* To smooth out random fluctuations.
Formula:
SMA_n = (\text{Actual Value}_1 + \text{Actual Value}_2 + ... + \text{Actual Value}_n) / n
Where n is the number of periods you are averaging.
Example/Application:
Let's forecast sales using a 3-month Simple Moving Average.
| Month | Actual Sales (₱) |
|---|---|
| Jan | 100,000 |
| Feb | 110,000 |
| Mar | 105,000 |
| Apr | 120,000 |
| May | 115,000 |
To forecast June's sales (using the last 3 months: March, April, May):
SMA_3 = (105,000 + 120,000 + 115,000) / 3
SMA_3 = 340,000 / 3
SMA_3 = \text{₱}113,333.33
So, the forecast for June sales is ₱113,333.33.
c. Weighted Moving Average (WMA)
Concept: Similar to SMA, but it gives more importance (weight) to more recent data points,
because recent information might be more relevant to the future. The sum of the weights
must equal 1 (or 100%).
In Simple Words: "The most recent stuff matters more than the older stuff."
When to Use:
* When there's a slight trend or you believe recent data is more indicative of the future.
Formula:
WMA = (\text{Weight}_1 \times \text{Actual Value}_1) + (\text{Weight}_2 \times \text{Actual
Value}_2) + ...
Example/Application:
Let's use the same sales data and forecast June sales using a 3-month Weighted Moving
Average with weights:
* May (most recent): 0.50 (50%)
* April: 0.30 (30%)
* March (oldest): 0.20 (20%)
(Notice: 0.50 + 0.30 + 0.20 = 1.00)
| Month | Actual Sales (₱) |
|---|---|
| Jan | 100,000 |
| Feb | 110,000 |
| Mar | 105,000 |
| Apr | 120,000 |
| May | 115,000 |
To forecast June's sales:
WMA = (0.50 \times 115,000) + (0.30 \times 120,000) + (0.20 \times 105,000)
WMA = 57,500 + 36,000 + 21,000
WMA = \text{₱}114,500
So, the forecast for June sales is ₱114,500.
d. Exponential Smoothing
Concept: This is a more sophisticated moving average that also gives more weight to recent
data, but it does so in a special way by "smoothing" the forecast over time. It requires only
the previous forecast, the most recent actual demand, and a "smoothing constant" (\alpha).
In Simple Words: "We'll adjust our old guess based on how much it was wrong last time."
When to Use:
* When there's no clear trend or seasonality.
* For continuous forecasting where you need to update forecasts frequently.
* When you need a simple method that requires little historical data storage.
Formula:
F_{t+1} = F_t + \alpha(A_t - F_t)
Where:
* F_{t+1} = Forecast for the next period
* F_t = Forecast for the current period
* A_t = Actual demand for the current period
* \alpha (alpha) = Smoothing constant (a value between 0 and 1)
What \alpha means:
* A higher \alpha (closer to 1) means the forecast reacts more quickly to recent changes in
actual demand. Use this if demand is volatile.
* A lower \alpha (closer to 0) means the forecast is slower to react to changes, resulting in a
smoother forecast. Use this if demand is very stable.
Example/Application:
Let's forecast sales for July using Exponential Smoothing.
* Actual sales for June (A_t) = ₱125,000
* Forecast for June (F_t) = ₱113,333 (from our Simple Moving Average example)
* Smoothing constant (\alpha) = 0.20
To forecast July's sales (F_{t+1}):
F_{July} = F_{June} + \alpha(A_{June} - F_{June})
F_{July} = 113,333 + 0.20(125,000 - 113,333)
F_{July} = 113,333 + 0.20(11,667)
F_{July} = 113,333 + 2,333.4
F_{July} = \text{₱}115,666.40
So, the forecast for July sales is approximately ₱115,666.
2. Causal Models / Associative Forecasting
Concept: Instead of just looking at past values of what you're trying to forecast (like sales),
causal models try to identify relationships between the variable you want to forecast (e.g.,
sales) and other external factors that might influence it (e.g., advertising spending, economic
growth, competitor actions).
In Simple Words: "What other things affect what I'm trying to predict?"
When to Use:
* When you believe there are measurable factors that drive the variable you want to
forecast.
* For medium to long-term forecasting.
The most common technique here is Regression Analysis.
a. Linear Regression
Concept: This method finds the "best fit" straight line through a set of data points to show the
relationship between two variables:
* Dependent Variable (Y): What you want to forecast (e.g., Sales).
* Independent Variable (X): The factor that you believe influences Y (e.g., Advertising
Spending, GDP, Time).
In Simple Words: "If I spend more on ads, how much more will I sell?"
When to Use:
* When you suspect a linear relationship between two variables.
* For trend forecasting (where X is time).
Formula for a Straight Line:
Y = a + bX
Where:
* Y = Dependent variable (e.g., Sales forecast)
* X = Independent variable (e.g., Advertising spending, or time period)
* a = Y-intercept (the value of Y when X is 0)
* b = Slope of the line (how much Y changes for every one-unit change in X)
Calculating 'a' and 'b': These formulas can be a bit complex by hand, but calculators and
spreadsheet software (like Excel) can do this easily.
* b = \frac{n(\sum XY) - (\sum X)(\sum Y)}{n(\sum X^2) - (\sum X)^2}
* a = \bar{Y} - b\bar{X} (where \bar{Y} is the average of Y, \bar{X} is the average of X)
Example/Application:
Let's say a company wants to predict sales based on their advertising spending.
| Month | Advertising Spending (X - in ₱1,000s) | Actual Sales (Y - in ₱1,000s) |
|---|---|---|
| 1 | 10 | 150 |
| 2 | 12 | 160 |
| 3 | 15 | 180 |
| 4 | 13 | 170 |
| 5 | 16 | 190 |
Using a spreadsheet program (like Excel's Data Analysis Toolpak or by using the formulas),
you might find:
* a = 90
*b=6
So the regression equation would be: Y = 90 + 6X
Now, to forecast sales if the company plans to spend ₱20,000 on advertising next month (X
= 20):
Y = 90 + 6(20)
Y = 90 + 120
Y = 210
So, the forecasted sales would be ₱210,000.
Understanding and Computing Forecast Error
No forecast is perfect! We need ways to measure how well our forecasts are doing. Forecast
error is simply the difference between what we predicted and what actually happened.
Formula:
Forecast Error = Actual Value - Forecasted Value
(Sometimes it's Forecasted Value - Actual Value; just be consistent)
Example:
* Actual Sales = ₱1,000,000
* Forecasted Sales = ₱1,200,000
Forecast Error = 1,000,000 - 1,200,000 = -\text{₱}200,000 (This means you over-forecasted
by ₱200,000).
Why measure error?
* To see if our forecasting method is working well.
* To compare different forecasting methods to find the best one.
* To understand the level of uncertainty in our predictions.
Important Concepts for Uncertainty in Forecasting
1. Standard Deviation (\sigma)
Concept: In forecasting, if you have a series of forecast errors (or actual values around a
mean), the standard deviation measures the spread or variability of those numbers around
their average.
In Simple Words: "How much do our forecasts (or actuals) typically differ from the average?
Is the forecast usually way off, or pretty close?"
* A small standard deviation means your actual outcomes tend to be very close to your
forecast (or the average forecast). High accuracy, low variability.
* A large standard deviation means your actual outcomes can be quite far from your
forecast. Low accuracy, high variability.
Application (Example):
If a forecast has a standard deviation of 50 units, it means that the actual results tend to
differ from the forecasted value by about 50 units on average.
2. PERT Analysis (Program Evaluation and Review Technique)
While PERT is commonly used for project management task durations, its principles can
apply to forecasting when there's significant uncertainty and historical data is limited.
Concept: PERT deals with uncertainty by asking for three estimates for a future event or task
duration:
* Optimistic (a): The best possible outcome (shortest time, lowest cost, highest sales) if
everything goes perfectly.
* Most Likely (m): The most probable outcome.
* Pessimistic (b): The worst possible outcome (longest time, highest cost, lowest sales) if
everything goes wrong.
From these three estimates, we can calculate:
a. Expected Value (E or EV)
Concept: This is a weighted average of the three estimates, giving more importance to the
"most likely" scenario. It's your best single guess for the outcome, considering the range of
possibilities.
In Simple Words: "What's our most realistic prediction, considering the best and worst
cases?"
Formula:
E = (a + 4m + b) / 6
Example/Application:
A new product's first month sales forecast:
* Optimistic (a) = ₱80,000
* Most Likely (m) = ₱100,000
* Pessimistic (b) = ₱160,000
E = (80,000 + 4 \times 100,000 + 160,000) / 6
E = (80,000 + 400,000 + 160,000) / 6
E = 640,000 / 6
E \approx \text{₱}106,667
So, the expected sales for the first month are approximately ₱106,667.
b. Standard Deviation (\sigma) in PERT
Concept: In PERT, the standard deviation tells you how much the actual outcome is likely to
vary from the expected value. A wider range between optimistic and pessimistic estimates
will result in a larger standard deviation.
In Simple Words: "How much wiggle room or uncertainty is there around our expected
value?"
Formula:
\sigma = (b - a) / 6
Example/Application (using the sales forecast above):
* Optimistic (a) = ₱80,000
* Pessimistic (b) = ₱160,000
\sigma = (160,000 - 80,000) / 6
\sigma = 80,000 / 6
\sigma \approx \text{₱}13,333
This means there's a variability of about ₱13,333 around the expected sales of ₱106,667.
3. Confidence Intervals
Concept: A confidence interval provides a range around your forecast (or expected value)
within which you are a certain percentage confident that the true actual value will fall. It
combines the expected value with its standard deviation.
In Simple Words: "We're X% sure that the actual result will be somewhere between Y and
Z."
Common Confidence Levels and Z-scores (assuming a normal distribution):
* 68% Confidence Interval: \pm 1 \times \sigma (Z-score \approx 1.0)
* 95% Confidence Interval: \pm 1.96 \times \sigma (Z-score \approx 1.96)
* 99% Confidence Interval: \pm 2.58 \times \sigma (Z-score \approx 2.58)
Formula for Confidence Interval:
\text{Confidence Interval} = \text{Expected Value} \pm (\text{Z-score} \times \text{Standard
Deviation})
Example/Application:
Using the PERT example, with Expected Value (E) \approx ₱106,667 and Standard
Deviation (\sigma) \approx ₱13,333.
To compute the 95% Confidence Interval for the sales forecast:
* Z-score for 95% = 1.96
Lower Bound = 106,667 - (1.96 \times 13,333)
Lower Bound = 106,667 - 26,132.68
Lower Bound \approx \text{₱}80,534
Upper Bound = 106,667 + (1.96 \times 13,333)
Upper Bound = 106,667 + 26,132.68
Upper Bound \approx \text{₱}132,800
So, based on the PERT estimates, the company is 95% confident that the actual first-month
sales will fall between approximately ₱80,534 and ₱132,800. This provides a much more
realistic picture than just a single expected value.
Conclusion
The Quantitative Approach to financial forecasting provides powerful tools to make
data-driven predictions. By understanding and applying techniques like Moving Averages,
Exponential Smoothing, Linear Regression, and the concepts of Standard Deviation and
Confidence Intervals (especially useful in PERT analysis), businesses can make more
informed decisions about their future operations, budgeting, and strategic planning.
While these methods are based on numbers, remember that they are still forecasts – they
give us the best estimate given past information, but the future can always hold surprises!