|
home | strategy development | strategy
skills | site
index | access
keys
Looking forward - Forecasting
>
in practice
Forecasting identifies and tracks past trends and
extrapolates them into the future. Typically, it is used to track over
time (time-series forecasting), and to make predictions about differences
among people, firms or other objects (cross-sectional forecasting). As
well as quantitative (statistical methods), it also includes the use of
more qualitative (judgmental) methods.
As looking into the future naturally involves a varying
degree of uncertainty, sometimes a distinction is drawn between
forecasting and projections. In certain contexts, particularly economic
ones, forecasting is used to refer to short-term extrapolations associated
with a reasonable degree of certainty. Projections are considered to be
longer-term, more sophisticated, but also less reliable. This distinction
does not always hold true, for example demographic projections can be very
reliable over the time span of a generation. For this reason this section
distinguishes instead between quantitative and qualitative trend analysis.
Quantitative Analysis
Quantitative trend analysis is probably the most common
forecasting method. It relies on the statistical analysis of historical
data - in other words it is relatively objective. Quantitative
techniques include extrapolation (such as moving averages, linear
projections against time or exponential smoothing) and econometric methods
(typically using regression techniques to estimate the effects of causal
variables). This type of analysis is commonly used to forecast demographic
and economic changes where extrapolating over time is believed to have
some validity.
The Strategy Unit, A Futurist's
Toolbox, identifies some of the main quantitative techniques used
by forecasters. Other techniques for short to medium term analysis and
forecasting include:
Modelling
Modelling is an extremely useful tool for quantitative
analysis. Excel and econometric modelling techniques are outlined in the modelling
section of the Guide.
Simple Moving Averages
The best-known forecasting method is moving averages.
It simply takes a certain number of past periods and adds them together,
then divide by the number of periods. Simple Moving Averages (MA) is an
effective and efficient provided the time series is stationary in both
mean and variance. The following formula is used in finding the moving
average of order n, MA(n) for a period t+1,
MAt+1 = [Dt + Dt-1 +
... +Dt-n+1] / n
where n is the number of observations used in the
calculation.
The forecast for time period t+ 1 is the forecast for
all future time periods. However, this forecast is revised only when new
data becomes available.
Weighted Moving Averages
Very powerful and economical. They are widely used
where repeated forecasts required-uses methods like sum-of-the-digits and
trend adjustment methods. As an example, a Weighted Moving Averages is:
Weighted MA(3) = w1.Dt + w2.Dt-1
+ w3.Dt-2
where the weights are any positive numbers such that:
w1 + w2 + w3 =1. A typical weights for this example is, w1 =
3/(1 + 2 + 3) = 3/6, w2 = 2/6, and w3 = 1/6.
Linear Projection
Used to estimate values in future periods. By taking
historical data, an actual growth rate can be determined. This rate is
then applied to the last known year and run forward. The validity of the
growth rate found in historical data depends largely on the number of
reference points and the period over which they are found. Obviously, the
more reference points and the longer the period, the better. Linear
projection will only serve as a predictor of future values if future trend
determinants are the same as historical determinants. Therefore, factors
such as technological innovation, changes in behaviour and radical
economic shifts can all mean that historical determinants are no guide to
future trends.
Often it is difficult to find sufficient data to allow
detailed quantitative analysis. Techniques to address this problem include
estimation and triangulation:
Estimation
One of the key difficulties in conducting forecasting
is a lack of available data. If this is the case, estimation may be
suitable. The most common forms of estimation are:
- Ask an expert or group of experts to use their experience to
formulate an opinion.
- Develop a case study. For example, how many gardens are there in the
UK? You might discover from the Office of National Statistics site
that there are x places of abode in the UK, of which b are units, c
are detached and d are terraces. You might assume that all the
detached and terrace properties have gardens and one third of the
flats have gardens. The most important thing is to ensure that your
assumptions are clearly noted, so that the model users are able to
adjust the assumptions if more accurate data comes to light.
- Mirroring. This method can be used when you identify a corresponding
event. For a particular prescription drug may always be bought in
conjunction with another drug. You may be able to ascertain the sales
of the second drug by adding up quantities from annual reports, and
then 'mirror' that number to find an estimation of the number of
sales of the first drug.
Triangulation
When developing a model, data is often incomplete or
approximate. In other instances you may have several sources of data that
conflict. One way of developing a base to work from is to triangulate the
available information to develop a defensible average.
Three sources of comparable data are needed. These may
be obtained by various methods - extrapolation, expert estimation, case
studies, literature reviews, etc. Once the information from all sources is
standardised (that is using the same base, units, denomination, etc), an
average is taken. Usually it is a straight average, though sometimes you
may weight some of the information sources - to reflect a higher quality
data source.
Qualitative Analysis
Qualitative trend analysis is more subjective and
is concerned mainly with social, institutional, commercial and political
themes (i.e. things which may not be linearly related to the past). For
example, qualitative trend analyses deal with issues such as:
- What is the future of trade unions?
- What is the future of political parties or NGOs?
- What is the future of the entertainment business?
One of the most common forms of qualitative trend
analysis is the identification of 'megatrends' - driving forces which
can change society in all spheres e.g. politics, economics, technology,
values and social relations. Other tools include scenarios
and analogies.
Qualitative analyses can be applied to most areas, but
work best when focusing on real change. Megatrends apply to all areas,
within the defined time and setting. It is important, though, to be aware
that mega-trends may themselves produce powerful counter-trends - and that
they may interact with each other.
Scenario Design
Quantitative and qualitative trend analyses together
form the basis for scenario
design. Different combinations of key trends are used to describe
possible pictures of the future, which can then be used to design or test
policy.
Strengths
- Quantitative forecasts are usually more objective, relatively
inexpensive and easy to use (contingent upon some knowledge of
statistics).
- Qualitative forecasts can be valuable predictors of new trends,
by using the creativity and good judgement of experts.
Weaknesses
- Quantitative forecasts can be misleading. The past is not always
predictive of the future. Such forecasts do not take into account
unpredictable changes or discoveries (e.g. discovery of new natural
resources) or 'wild cards' (e.g. unexpected acts of terrorism).
- When using qualitative techniques to identify possible new
trends it will always be the case that some, or maybe even all of the
results are eventually disproved. It is particularly difficult to
distinguish between short term 'fads' and long term trends.
References
The Strategy Unit report A
Futurist's toolbox sets out the basic steps for carrying out
forecasting analysis. The report summarises the six key methodologies for
futures work, covering most of the commonly used tools by professional
futurists. Some of the elements of the report are outlined below.
Short
Survey of Published Material on Key UK Trends 2001-2011 This report
was undertaken by the Strategy Unit with the Defence Science and
Technology Laboratory (DSTL) to synthesise existing predictions on trends
in several sectors including: the economy; demographics; the environment;
housing; and, health. The data is broken down by time into a period of
relative certainty (2001-2006) and a period of lower certainty
(2006-2011).
Strategic
Futures Thinking: meta analysis of published material on Drivers and
Trends. This was another report produced in conjunction with the
Defence Science and Technology Laboratory (DSTL). The report examines
published literature on key futures issues. It identifies six key drivers
of change; demographics, economics, science and technology, environment,
governance and attitudes and beliefs. It is also a useful source document
for other materials.
The OECD
International Futures Programme. This is designed to help decision
makers to understand the key factors affecting the long-term future. It
provides monitoring of the long-term economic and social horizon. It also
provides early warning on emerging issues, pinpointing of major
developments and possible trend breaks.
For comprehensive information on all aspects of forecasting from
methods to purposes to evaluation there is a useful website - the Forecasting
Principles site run by Wharton Business
School. The work outlines a number of different ways to approach
forecasting and provides a forecasting methodology tree for determining
which forecasting method is most appropriate. While Strategy Unit cannot
vouch for the day to day currency of this site, at the time of writing it
summarises much useful knowledge about forecasting. It is designed to be
accessible to researchers, practitioners, and educators. This knowledge is
provided as principles (guidelines, prescription, rules, conditions,
action statements, or advice about what to do in given situations). There
are many materials that can be downloaded.
Forecasting
In Practice 1: SU Ethnic Minorities and the Labour Market Project
This project was set up to examine and improve the
position of ethnic minorities in the UK labour market. As a first stage,
an Interim report was produced in early 2002. Amongst other things, the
interim report looked at the future size of the ethnic minority population
within the UK, as well as the effects that this would have on the labour
market as a whole. The project did not do it's own forecasting, but
rather used existing forecasts produced in this area. This data is
discussed on page 24 of the Ethnic Minorities and
the Labour Market report.
The forecasts used were important in stressing the fact
that the problems faced by ethnic minorities in the labour market are
growing to a point at which they become a problem for the wider UK
population. In other words, forecasting helped to show how a niche concern
is likely to become a general one over time.
|
Forecasting
In practice 2: SU Waste Project
The Waste Project utilised a linear projection model that projected, on
an annual basis to 2020, volumes of municipal waste and waste management
methods, volumes of recyclate, expenditure and facilities.
Prediction of the volume of arisings for the entire
period was therefore a crucial part of the model. However, the extent to
which linear projection was used in the Waste Project, to estimate growth
rates of municipal waste, was limited due to lack of historical growth
data. Detailed data was not collected until about 5 years before the
study, and even that data was incomplete. Furthermore the data that did
exist was controversial - industry sources questioned whether the
numbers reported related solely to the stream in question, given the
unrecorded transfers between, for example, municipal and commercial waste.
To add further complication, there was no consensus over the growth
drivers or trends, making linear projection difficult to do and defend as
the sole method.
This problem was partly resolved by using linear
projection, in conjunction with estimation, to run two different growth
rates on top of each other. Firstly, a generic 3% growth rate, based on
growth in the previous period, was used, with the default growth rate
becoming 2.5% from 2010 onwards. Secondly, a set of waste minimisation
programs in the scenario necessitated a separate growth rate for specific
targeted materials in the waste stream, hence, a more complex series of
estimations, which were not based on historical data, were overlaid the
generic growth rate. These estimates were forward looking and based on how
waste minimisation programs, e.g. reducing household waste through
producer responsibility, home composting, disposable nappy reduction etc,
would further reduce selected material streams. The growth rates resulting
from the waste minimisation program were determined using a variety of
sources.
In a modelling situation where there is uncertain data,
or where more information is likely to emerge over time, which will alter
the growth rate and/or increase the confidence of the estimation, it is
useful to allow the model user to be able to change the questionable
variable. The model must then be correctly linked to the variable data to
be able to reflect such changes.
|
^ top
|
