ShoTS Forecasting: Short Time Series Forecasting for Management Research
| Published date | 01 April 2023 |
| Author | Dimitrios Thomakos,Geoffrey Wood,Marilou Ioakimidis,Giorgos Papagiannakis |
| Date | 01 April 2023 |
| DOI | http://doi.org/10.1111/1467-8551.12624 |
British Journal of Management, Vol. 34, 539–554 (2023)
DOI: 10.1111/1467-8551.12624
ShoTS Forecasting: Short Time Series
Forecasting for Management Research
Dimitrios Thomakos ,1Geoffrey Wood,2,3 Marilou Ioakimidis4,5
and Giorgos Papagiannakis4
1Department of Business Administration, National and Kapodistrian University of Athens, Athens, 10559,
Greece, 2DAN Management and Organizational Studies, Western University, London, Ontario, N6A 3K7,
Canada, 3Trinity College Dublin, Universityof Dublin, Dublin 2, D02 PN40, Ireland, 4Department of
Economics, University of the Peloponnese, Tripoli, 22100, Greece, and 5Department of Economics, National
and Kapodistrian University of Athens, Athens, 10559, Greece
Corresponding author email: dthomakos@ba.uoa.gr
Wepresent a novel method for forecasting with limited information, that is for forecasting
short time series. Our method is simple and intuitive; it relates to the most fundamental
forecasting benchmark and is straightforward to implement. We present the technical
details of the method and explain the nuances of how it works via two illustrative exam-
ples, with the use of employment-related data. We nd that our new method outperforms
standard forecasting methods and thus offers considerable utility in applied management
research. The implications of our ndings suggest that forecasting short time series, of
which one can nd many examples in business and management, is viable and can be of
considerable practical help for both research and practice – even when the information
available to analysts and decision-makers is limited.
Introduction
Forecasting in business is a vital activity for plan-
ning and strategizing, one that occurs in virtu-
ally every area, including marketing (Makridakis
and Wheelwright, 1977), operations (Fildes et al.,
2008), HRM (Malik, 2018; Turner, 2002), start-
ups (Hyytinen, Lahtoneni and Pajarinen, 2014),
diffusion of innovations (Meade and Islam, 2006),
nance (Lin, Wu and Zhou, 2018) and business
protability (Ding, Zhang and Duygun,2019). In-
deed, forecasting is universally accepted by man-
agement as a necessary business function (Drury,
1990; Hogarth and Makridakis, 1981). Neverthe-
less, there is a considerable gap between the us-
age of forecasting by management practitioners
and researchers. Withinbusiness and management
studies, and outside of a few specialized journals,
[Correction added on 14 June 2022, after rst online pub-
lication: In text equation on page 6 has been updated in
this version.]
there is limited research on or using forecasting,
with the most salient work being encountered in
operations management (cf. Ren et al., 2020) and
marketing (Kumar et al., 2020; Makridakis and
Wheelwright, 1977). This could be because the
common ways of using forecasting may not be
enough to withstand the rigours of peer review,
the relevance of established forecasting tools for
theory-informed research questions, or simply be-
cause it is one area where advancesin methodology
have yet to percolate across the wider business and
management research community.
In making use of forecasting in business and
management research, key issues to be considered
are the variable(s) to be predicted, the accuracy
required, the horizon and timing of the forecast
and, more importantly,the data on which the fore-
cast is based (Makridakis and Wheelwright, 1977;
Ren et al., 2020). Accuracy is a key dimension of
a forecast; an inaccurate prediction can easily re-
sult in losses for a business (Mizan and Taghipour,
2021), and this may also lead to research lacking in
A free video abstract to accompany this article can be foundonline at: https://www.youtube.com/watch?v=T0nj-eztThI
A free Teaching and Learning Guide to accompany this article is available at: http://onlinelibrary.wiley.com/journal/
10.1111/(ISSN)1467-8551/homepage/teaching___learning_guides.htm
© 2022 British Academy of Management.
540 D. Thomakos et al.
rigour. In turn, the accuracy of quantitative busi-
ness forecasts depends on the availability of rele-
vant data. When time series data arelimited, it be-
comes more difcult to detect any time series pat-
tern, t an appropriate model and generate an ac-
curate forecast. Moreover, in such instances, each
data point becomes much more important than if
the data were more comprehensive, which can lead
to biased forecasting results (Weigand, Lange and
Rauschenberger, 2021). In some cases, there may
be no clear trend shown by a short time series,
whereas a pattern might be revealed by more com-
prehensive data (Mizan and Taghipour, 2021).
Yet, in a fast-changing world, with regular
system-challenging events, the range of usable
and/or relevant past observations is often limited
(Goyal and Karande, 2021). In practice, there is
often a need to formulate a forecastfor a decision-
making process using a relatively small number of
observations. For example, limited data occur in
the cases of early-stage entrepreneurship (Hyyti-
nen, Lahtoneni and Pajarinen, 2014), expected
sales of new products or the rate of diffusion
of a new technology (Christodoulos, Michalake-
lis and Varoutas, 2010); there may also be trends
taking place during a particular ‘exceptional time’
such as a pandemic. Also, it may be difcult to
predict future trends based on evidence before a
watershed event (e.g. a major nancial crisis).
Moreover, sectoral dynamics may restrict the ap-
plicability or relevance of a forecast. In the world
of fast fashion, for example, manufacturersand re-
tailers nd it difcult to predict the fortunesof new
products based on data frompast ones (Choi et al.,
2014). Yet, whenever a business is new or is ex-
panding into a fresh venture or product,or in times
of socio-economic change or instability (Wieland,
2021), those wishing to make use of forecasting
have to work with short time series with limited
data.
Even when there is more data, many databases
that are used by management researchers are sub-
scribed to by their libraries or via grant funding
for xed periods; this is particularlythe case within
institutions and national higher education systems
undergoing austerity. Other well-known databases
(e.g. Zephyr)may have incomplete data on specic
variables of interest to the researcher. Forecast-
ing in these conditions or, indeed, in any situation
where key variables have been poorlyrecorded and
the time series is abbreviated is particularly chal-
lenging. Forecasting models that work well when
there are many observations in a series may be of
very limited utility for short time series with sparse
data. Clearly, there is a need for robustmethodolo-
gies for forecasting short time series.
Given the above considerations, we present a
novel forecasting model to be used with short time
series with limited data, which we correspond-
ingly term ShoTS, forshort time series forecasting.
This model could therefore be used when obser-
vations on a variable of interest are relatively few,
yet where there is value in deriving a reasonable
forecast.
The paper is divided into ve sections,including
this introduction. The second section consists of
a review of the literature, including other research
on forecasting with limited data. In the third sec-
tion, we present our approach, providing a step-
by-step presentation of the process along with ba-
sic equations. We also explain the model’s main
advantage in comparison to other forecasting
methods. In the fourth section, we provide two
illustrative examples using the new model, again
comparing the model to other methods. A discus-
sion in which we detail the economic signicance
of the paper follows in the nal section.
Literature review
Forecasting in management research and
applications
There are two broad types of forecasting ap-
proaches. Qualitative approaches generally base
a prediction on the views of a panel of experts.
Quantitative approaches make use of data gath-
ered to forecast a future quantity or quantities of
interest (for a thorough review of various fore-
casting technics, see Petropoulos et al., 2022).
For quantitative approaches, the time horizon of
the forecasting (e.g. forecast for the next month
or year) is important in choosing the forecasting
methodology. This is because an approach that
typically has good results for short-term predic-
tions may provide poor results for a medium-
or long-term forecast, and vice versa (Chambers,
Mullick and Smith, 1971). The simplest of the
quantitative forecastingmethodologies is the naïve
forecast, which bases the forecast on the last ob-
servation made in a time series, and this bench-
mark is the point of departure for our proposed
method. More complex, but still easily accessi-
ble, methods include the exponential smoothing
© 2022 British Academy of Management.
Get this document and AI-powered insights with a free trial of vLex and Vincent AI
Get Started for FreeStart Your Free Trial of vLex and Vincent AI, Your Precision-Engineered Legal Assistant
-
Access comprehensive legal content with no limitations across vLex's unparalleled global legal database
-
Build stronger arguments with verified citations and CERT citator that tracks case history and precedential strength
-
Transform your legal research from hours to minutes with Vincent AI's intelligent search and analysis capabilities
-
Elevate your practice by focusing your expertise where it matters most while Vincent handles the heavy lifting
Start Your Free Trial of vLex and Vincent AI, Your Precision-Engineered Legal Assistant
-
Access comprehensive legal content with no limitations across vLex's unparalleled global legal database
-
Build stronger arguments with verified citations and CERT citator that tracks case history and precedential strength
-
Transform your legal research from hours to minutes with Vincent AI's intelligent search and analysis capabilities
-
Elevate your practice by focusing your expertise where it matters most while Vincent handles the heavy lifting
Start Your Free Trial of vLex and Vincent AI, Your Precision-Engineered Legal Assistant
-
Access comprehensive legal content with no limitations across vLex's unparalleled global legal database
-
Build stronger arguments with verified citations and CERT citator that tracks case history and precedential strength
-
Transform your legal research from hours to minutes with Vincent AI's intelligent search and analysis capabilities
-
Elevate your practice by focusing your expertise where it matters most while Vincent handles the heavy lifting
Start Your Free Trial of vLex and Vincent AI, Your Precision-Engineered Legal Assistant
-
Access comprehensive legal content with no limitations across vLex's unparalleled global legal database
-
Build stronger arguments with verified citations and CERT citator that tracks case history and precedential strength
-
Transform your legal research from hours to minutes with Vincent AI's intelligent search and analysis capabilities
-
Elevate your practice by focusing your expertise where it matters most while Vincent handles the heavy lifting
Start Your Free Trial of vLex and Vincent AI, Your Precision-Engineered Legal Assistant
-
Access comprehensive legal content with no limitations across vLex's unparalleled global legal database
-
Build stronger arguments with verified citations and CERT citator that tracks case history and precedential strength
-
Transform your legal research from hours to minutes with Vincent AI's intelligent search and analysis capabilities
-
Elevate your practice by focusing your expertise where it matters most while Vincent handles the heavy lifting
