Unit 1: The essentials of forecasting
Grado Administración y Dirección de Empresas
Business Workshop II
Juan A. Román Aso
Year: 2024/2025
Adapted from Hyndman and Athanasopoulos (2018): Forecasting:
Principles and Practice, 2nd Edition, Otexts
UNIT 1
1.1 Introductory issues
1.1 Introductory issues
The economic science studies how the economic agents
interact in competitive markets and the macro
consequences of these individual choices.
When a family chooses a new car, when a company takes
on staff or when the government increases public
spending, they are making choices with unknown
consequences.
1.1 Introductory issues
• If the family would know that the car engine is going to
be a source of problems, they wouldn’t buy the car.
• If the company would know who is the best employee
among the candidates, the choice would be easy
• If the government would anticipate the effect of any
political measure, the differences between parties
would be insignificant.
1.1 Introductory issues
However, economic agents often interact in a context of
uncertainty. Therefore, there is a need for forecasting to
cover the risks associated with poor choices. Forecasting
techniques provide economic agents with the best
instruments to decision making process in a context of
global uncertainty.
1.1 Introductory issues
Chambers, Mullick y Smith (1971): “In virtually every
decision they make, executives today consider some kind
of forecast. Sound predictions of demands and trends are
no longer luxury items, but a necessity, if managers are to
cope with seasonality, sudden changes in demand levels,
price-cutting maneuvers of the competition, strikes, and
large swings of the economy. Forecasting can help them
deal with these troubles”
1.1 Introductory issues
The public officers, business managers and
families are daily facing dilemmas about
economic issues. A simple choice could
generate a profit or a loss
1.1 Introductory issues
Not all the variables have the same predictability. It
depends on two factors:
1. The availability of past data and a reliable method to
predict.
2. The number of factors that affects to the evolution of
the variable
1.1 Introductory issues
If we want to forecast the number of foreign tourists in
Spain, we have enough past data, we know how the
accurate method is and what are the most important
factors (per capita income of foreign countries, inflation or
political situation in competence markets).
However, if I could predict the
lottery number , Do you think that I
would be here?
1.1 Introductory issues
How can we determine the goodness of a forecast?
Essentially, better forecasts are able to capture the
behavior patterns in the historical data and the
relationships with other variables.
In addition, a good forecast takes into consideration the
changing environment, the volatility and the fluctuations
Exercise 1
Write a brief report (no
Read the article on
https://www.bizeducator.c more than 250 words)
om/the-importance-of-fore
about the growing
casting-in-business/
and watch the video importance of forecasting
included.
UNIT 1
1.2 Key concepts
1.2 Key concepts
Forecasting is commonly implemented in business as it
benefits the decision-making process about scheduling of
production, transportation and personnel, and provides a
guide to long-term strategic planning. Nevertheless,
forecasting is often confused with planning or goals. Here
you have the main differences:
1.2 Key concepts
• Forecasting permits us to predict the future in the best possible
way. We use historical data and knowledge of future events
• Goals are what you would like to happen. Must be linked to
forecasts and plans
• Planning is the way to response to forecasts and goals. Suitable
actions required to make your forecast match your goals.
1.2 Key concepts
• Short-term forecasts: are needed for the scheduling of personnel,
production and transportation. Forecast of demand too
• Medium-term forecasts: are needed to determine future
resource requirements, that is, to buy raw materials, machinery,
equipment… or hire personnel
• Long-term forecasts: are used in strategic planning. For instance,
market opportunities or env factors
1.2 Key concepts
Based on past observations, forecasting provides
quantitative estimates of the future evolution of a
variable.
UNIT 1
1.3 Basic steps
1.3 Basic steps
1. Define the problem and the objective
In this step, we should understand how forecasting can
help you in this task but being aware that it does not
provide a perfect response, the result is just an
estimation. It is also useful to think about who requires
the forecasts (organisation, public officers…)
1.3 Basic steps
2. Collect the data
In this step, we gather the information needed to conduct
the forecast. On one side, you need statistical data and on
the other skilled, and experienced workers to collect the
data and use the forecasts.
1.3 Basic steps
3. Exploratory study
After having reviewed the data, you must determine if the
time series presents trend, seasonal component, structural
breaks or outliers.
1.3 Basic steps
4. Select the suitable technique.
The availability of past data and the existence of significant
relationships between the forecast variable and other
factors determine the approach selected. In this course, we
will study the most important techniques that are currently
being used to forecast in the business environment.
1.3 Basic steps
5. Evaluate the forecast.
In this step, we evaluate the quality of the forecasting
conducted in order to know if the method selected was
accurate or otherwise, we must select another one
Exercise 2
Choose one of the following options and explain
briefly how to apply the five steps in each case:
1. Forecast the electricity demand
2. Forecast the Exchange rate €/$
3. Forecast the public spending in health
4. Forecast the sales of IBM
1.4
Sources of problems
1.4 Sources of problems
Are forecast really useful? Before the onset of the Great
Recession (2009) many forecasters and institutions
committed to perform forecasts of economic variables had
predicted an improvement in world economic growth. The
collapse of Lehman Brothers on September 15 th, 2008, and
the subsequent crisis affected all the forecasts proposed
before, bounding considerably their validity. Why did it
happen? Unexpected change in the environment.
1.4 Sources of problems
If available data are robust, the model selected to perform
the forecast is adequate and the environment is stable,
the forecast will be close to real value and the error will be
insignificant. But this panorama does not happen very
often.
We can find the following sources of error:
1.4 Sources of problems
1. Structural shifts:
This source of error takes place when an external factor
unexpectedly changes. Suppose you run a bakery in a small
town and suddenly, Panishop opens a store in the same
town, two streets away. Obviously, your sales will suffer a
remarkable drop
1.4 Sources of problems
2. Model misspecification
Suppose you are trying to determine the factors that affect
to the public spending in health for Spain. You propose a
structural model where explanatory variables are:
• Tax revenues
• Population
• Population under 14 What variable
is lacking?
• Pollution
• Number of smokers
1.4 Sources of problems
3. Inaccurate data
Macroeconomic data (GDP, Inflation are usually published by
public institutions: International Monetary Fund, World
Bank, Eurostat or the National Statistics Institute, among
others. Financial data on stocks and companies are
published by corporations.
Many databases are created with the information obtained
from surveys.
1.4 Sources of problems
3. Inaccurate data
Sampling always involves an unavoidable error
although the process would have been
developed by expertise people
In the book Practical Business Forecasting
(2003), Evans states:
1.4 Sources of problems
3. Inaccurate data
“The decennial census is supposed to count
every person in the US, but statisticians
generally agree the reported number of people
in large cities is significantly less than the
actual number…Thus even in this most
comprehensive data collection effort, which is
supposed to count everyone, some errors
remain. It is reasonable to assume that errors
1.4 Sources of problems
4. Policy shifts
Changes in monetary or fiscal policies can derive in
unexpected fluctuations as described in Evans (2003):
“Anyone who tries to estimate an equation to predict short-
term interest rates will soon find that, during the mid-
1970s, Fed Chairman Arthur Burns used monetary policy to
offset the recessionary impact of higher oil prices, leading
to unusually low real interest rates…”
1.4 Sources of problems
4. Policy shifts
…whereas in the early 1980s, Chairman Paul Volcker
refused to accommodate the further increase in oil prices,
leading to unusually high real interest rates. The real
Federal funds rate equals the nominal rate minus the
change in inflation over the past year”
1.4 Sources of problems
5. Changes in expectations
To explain the changes in expectation, let´s see a fragment
of Evans (2003):
“I have already noted how the Fed acted quite differently to
the first and second energy shocks in 1973–4 and 1979–80
respectively. However, that was not the only change; private
sector economic agents also reacted differently…
1.4 Sources of problems
5. Changes in expectations
…The first energy shock was viewed by most consumers
and businesses as a once-in-a-lifetime event, so they did
not alter their behavior patterns very much...
1.4 Sources of problems
5. Changes in expectations
…As a result, oil imports continued to increase, and
eventually oil prices rose again. After the second energy
shock, attitudes changed significantly. Most people now
expected that massive price increases would continue on a
regular basis, and forecasts were common that oil prices
would rise to $100/bbl by the end of the twentieth
century….
1.4 Sources of problems
5. Changes in expectations
…both consumers and businesses started using less energy,
buying more fuel-efficient motor vehicles, and constructing
more fuel-efficient buildings…. Any forecast of the economy
in the 1980s – whether right or wrong – was influenced by
the assumption about energy prices”.
Exercise 3
Analyse how the problems described above can affect to
your forecast of :
1. Energy demand
2. Earnings of the banking sector
1.5
Types of forecasts
1.5.1.Point or interval
When we talk about forecasting, we usually refer to point
forecast:
The sales in t+1 will reach 100,000€
However, for decision making process, it may be more
suitable an interval forecasting:
The sales will range between 90,000 and 110,000
1.5.1. Point or interval
When do we need an interval?
“Suppose a company has a limited amount of excess
manufacturing capacity. If sales grow less than 5% per
year, the company will be better off using its existing
facilities. If sales grow more than 5% per year, it will be
better off building a new plant. In this case, the point
estimate for sales growth is not as important as the
probability that sales growth will exceed 5%...
1.5.1. Point or interval
A similar case might be made for advertising budgets. If a
firm thinks a $1 million expenditure on advertising will
boost sales by at least $5 million, it will decide to go ahead
and spend the money. It doesn’t matter so much whether
the increase in sales is $6 or $10 million, but if it is $4
million, the expenditure will not be made.
1.5.1. Point or interval
…At the macro level, suppose the Fed decides that 3% is the
highest level of inflation that is tolerable. If inflation rises
1%, 1/2%, or 2%, there will be no change in monetary
policy. If it exceeds 3% – or if it appears likely it will soon
exceed 3% if policy is not changed – the Fed will boost
short-term interest rates”. Evans (2013)
1.5.2.Absolute or conditional
Absolute: The inflation will rise to 4%
Conditional: the inflation will rise to 4% if public spending
increases 3%
Suppose you are a speculator that operates in financial
markets, what do you prefer to know?
Obviously you prefer a)
a) If prices will rise or fall
b) If prices will rise or fall under certain circumstances
1.5.3 Alternative scenarios weighed by
probabilities
In many cases, economic agents make choices according to
a specific scenario. Suppose you run a company that makes
a profit if the economic growth is positive and otherwise, it
makes a loss. The best procedure is evaluating the
probability of a recession the next year
1.5.3 Alternative scenarios weighed by Example
probabilities
Consider a business manager that has to decide between
two options (D1 and D2) with uncertain consequences.
The gain will depend on two events (A1 and A2). The
business manager does not know how likely is each one.
1.5.3 Alternative scenarios weighed by Example
probabilities
However, business manager knows the potential gains of
having each decision:
A1 A2
D1 10 30
D2 0 40
1.5.3 Alternative scenarios weighed by Example
probabilities
Then, consider that business manager conducts a
probability distribution for A1 and A2 based on historical
data.
P(A1) = 0,3 The 30% of times, Event A1 occurs; the 70% of
times, the event A2 occurs
P(A2) = 0,7
1.5.3 Alternative scenarios weighed by Example
probabilities
If the manager would have perfect information (she/he
knows what event is going to occur), the right choice will
be always made and the expected gain is
G=0,3*10+0,7*40= 31 A1 A2
D1 10 30
D2 0 40
1.5.3 Alternative scenarios weighed by Example
probabilities
As this scenario is not realistic and the uncertainty is
measured with probability, let us see what is happening:
• When the event A1 is going to occur, the manager uses
the imperfect information available to create this
probability distribution:
p(A1)=0,9; p(A2)=0,1.
Manager senses that A1 is going to happen, but there is not perfect
information, then what does manager do?
1.5.3 Alternative scenarios weighed by Example
probabilities
In consequence, when event A1 is going to occur, the
manager compares the gains of making D1 and D2:
10*0,9+30*0,1 = 12 > 0*0,9+40*0,1 = 4.
If manager makes decision If manager makes decision
1, the gain is 12 2, the gain is 4
A1 A2
D1 10 30 Manager
D2 0 40 makes D1
1.5.3 Alternative scenarios weighed by Example
probabilities
• When the event A2 is going to take place, the manager
uses the imperfect information available to create this
probability distribution:
p(A1) =0,2 ; p(A2) =0,8.
Manager senses that A2 is going to happen, but there is not perfect
information, then what does manager do?
1.5.3 Alternative scenarios weighed by Example
probabilities
The manager compares the gains of making D1 and D2
10*0,2+30*0,8 = 26 < 0*0,2+40*0,8 = 32.
If manager makes decision If manager makes decision
1, the gain is 26 2, the gain is 32
A1 A2
D1 10 30 Manager
D2 0 40 makes D2
1.5.3 Alternative scenarios weighed by Example
probabilities
Then, if the manager makes decision 1 when the event A1
occurs and makes D2 when A2 occurs, we can confirm that
manager uses the imperfect information to make a right
choice.
The gain is the same to that obtained by assuming perfect
information
10*0,3+40*0,7 = 31.
Exercise 4
Consider the gains presented above but the information
available leads manager to develop these probability
distributions:
When A1 takes place When A2 takes place
p(A1)=0,4; p(A2)=0,6. p(A1) =0,5 ; p(A2) =0,5.
What would the manager do in each case?
Bibliography
Hyndman and Athanasopoulos (2018): Forecasting:
Principles and Practice, 2nd Edition, Otexts