Forecasting Errors of New Venture Survival

• Published date: 01 December 2014
• DOI: http://doi.org/10.1002/sej.1187
• Date: 01 December 2014

FORECASTING ERRORS OF NEW VENTURE

SURVIVAL

ARI HYYTINEN1,2*, JUKKA LAHTONEN1, and MIKA PAJARINEN3

1School of Business and Economics,University of Jyväskylä, Jyväskylä, Finland

2Yrjö Jahnsson Foundation, Helsinki, Finland

3The Research Institute of the Finnish Economy (ETLA), Helsinki, Finland

This article studies entrepreneurs’ forecast errors around market entry. Using data on nascent

entrepreneurs in the U.S. and start-ups in Finland, we find that besides being overoptimisticon

average in both countries, entrepreneurs’ survival expectations can barelydistinguish survival

from exits. Moreover, about one fourth of the entrepreneurs do not provide an estimate for the

survival of a typical venture. However, among those that do provide it, the estimates are less

overoptimistic. We also compare the forecast accuracy of entrepreneurs to those of macroeco-

nomic forecasters. Our findings provide guidance for the development of positive theories of

entrepreneurial belief formation and overoptimism. Copyright © 2014 Strategic Management

Society.

INTRODUCTION

Recent research both in economics (Manski, 2004)

and entrepreneurship (Shane and Venkataran, 2000;

Cassar, 2006; McMullen and Shepherd, 2006;

Shepherd, McMullen, and Jennings, 2007; Felin and

Zenger, 2009; Dimov, 2010) emphasizes the impor-

tance of understanding better the effects of subjec-

tive expectations (beliefs) on behavior, as well as

their interaction. McCann and Folta (2012) demon-

strate, for example, why expectations data are

needed to explore the thresholds of entry into the

entrepreneurial process. McCann and Vroom (2013)

study, in turn, how nascent entrepreneurs revisetheir

beliefs during the entry process and consider its

implications for, for example, opportunity evaluation

and exploitation.

A particularly prominent finding in the prior lit-

erature is that entrepreneurs are prone to be overly

optimistic around and after the market entry

(Cooper, Woo, and Dunkelberg, 1988; Landier and

Thesmar, 2009, and Cassar, 2010, 2012).1Such over-

optimism is present if the (potential) founders of new

businesses overestimate the chances of success or

development of their business compared to its actual

longevity and performance. Building on this,

Hayward, Shepherd, and Griffin (2006) have put

forward a hubris theory of entrepreneurship to

explain why so many entrepreneurs choose to enter

and yet subsequently fail (see also Dushnitsky, 2010;

Camerer and Lovallo, 1999). A number of formal

economic models also resort to various notions of

entrepreneurial overconfidence and overoptimism

(de Meza and Southey, 1996; Manove and Padilla,

1999; Hyytinen, 2003; Coval and Thakor, 2005).

Keywords: entrepreneurship; survival; expectations; overopti-

mism; forecast error

*Correspondence to: Ari Hyytinen, School of Business and

Economics, University of Jyväskylä, P.O. Box 35, Jyväskylä,

Finland, 40014. E-mail: ari.hyytinen@jyu.fi.

1Other often cited evidence on entrepreneurial overoptimism

measures it indirectly using data on respondents’ expectations

about general economic outcomes or life events (Arabsheibani

et al., 2000; Puri and Robinson, 2007). However, not all

studies find evidence for overoptimism. For example, Lowe

and Ziedonis (2006) are not able to link entrepreneurial

overoptimism with the commercialization efforts of U.S. uni-

versity inventions. See also Hogarth and Karelaia (2012), who

question the view that the process of entry is associated with

overoptimism.

bs_bs_banner

Strategic Entrepreneurship Journal

Strat. Entrepreneurship J., 8: 283–302 (2014)

Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/sej.1187

Copyright © 2014 Strategic Management Society

As defined, entrepreneurial overoptimism is about

entrepreneurs overestimating the chances of success

of their business compared to its objective perfor-

mance.2Therefore, it refers to a systematic positive

bias in the forecasts relative to the objective likeli-

hood of an event. This implies, in turn, that there are

systematic patterns in the forecast errors of entrepre-

neurs. What the prior literature has not, however,

recognized is that there are two dimensions to entre-

preneurs’ forecast errors: first, expectations can be

miscalibrated relative to the objective likelihood of

an event. For example, a set of probabilistic forecasts

(expectations) for a binary event are not well cali-

brated if, on average, they do not match with the

observed relative frequencies of the event. So far,

entrepreneurial forecast errors have typically been

understood in terms of overoptimism and, thus,

miscalibration. However, expectations may also

have weak resolution. This is a second dimension of

forecast errors, and it is distinct from the (potential)

miscalibration of forecasts: a set of forecasts have

poor resolution if the event of interest is rarely pre-

dicted to happen with a high probability when it

really happens and if the event is rarely predicted to

happen with a low probability when it does not

happen.3As we explain in detail later, such fore-

casts contain a limited amount of discriminating

predictive information because the degree to which

different subjective probabilities (expectations) are

followed by the correct realizations is low. In con-

trast, a set of forecasts have good resolution if the

large (small) probabilistic forecasts of an event are

regularly followed by the event happening (not

happening). In this case, the forecasts are strongly

correlated with the realizations and they allow

one to sort the realizations into two groups that differ

in terms of the relative frequency of the event

happening.4

Both the calibration of forecasts and their resolu-

tion power are scrutinized routinely in the forecast-

ing literature (see Diebold and Rudebusch, 1989;

Galbraith and van Norden, 2011, 2012; Lahiri and

Wang, 2013). This article argues that they are also

useful for understanding nascent entrepreneurship

and entry process.

First, the nature of entrepreneurial beliefs, the

associated forecasts errors, and how the (initially

erroneous) beliefs are revised appear to be an inte-

gral element of the entire nascent entrepreneurial

process because they affect opportunity identifica-

tion, evaluation, and exploitation (Shepherd et al.,

2007; McCann and Vroom, 2013). The accuracy of

entrepreneurs’ beliefs is also decisive for creation of

competence-based resource advantages (Barney,

1986; Makadok, 2003). How those beliefs should be

formed and are revised depend on the nature of fore-

casts errors. For example, increasing the awareness

of potential entrepreneurs about base rates, i.e., the

average survival likelihood of similar new ventures,

is not likely to increase the resolution power of

the entrepreneurs’ (survival) forecasts, but could

increase calibration.

Second, the prior forecasting literature has long

recognized that being able to correctly identify cases

in which the probability of the event is extraordi-

narily low or high is particularly useful for decision

making (Murphy, 1973; Diebold and Rudebusch,

1989; Lahiri and Wang, 2013). This refers to fore-

casts’ resolution and is obviously valuable for oppor-

tunity evaluation and exploitation and the associated

series of decisions that nascent entrepreneurs have to

make. Better resolution power may reduce, for

example, the number of missed opportunities, i.e.,

the number of people who fail to enter due to

2The terminology is somewhat unsettled here. In Moore and

Healy (2008), overestimation refers to an inaccurate forecast

relative to the (objective) likelihood of an event. Our analysis

bears on such inaccuracies, but we use the term overoptimism

instead. Moore and Healy (2008) also consider two other types

of overconfidence: first, overplacement, which corresponds to

the better-than-average -effect (i.e., interpersonal optimism);

and second, overprecision, which is about the tendency of

people to be overly confident about how accurate their forecasts

or expectations are. Examples of the papers in entrepreneurship

research that focus on overprecision in estimation and

overplacement include but are not limited to Forbes (2005) and

Ucbasaran et al. Camerer and Lovallo (1999: 306) talk in their

seminal paper about ‘overconfidence’ when theyrefer to ‘man-

agers acting on the optimism about relative skill they

exhibit. . .’

3The precise term that is used to describe this dimension of

forecast errors varies a bit, but we use the resolution term, as it

has been used in the recent economic forecasting literature; see

Lahiri and Wang (2013) and Galbraith and van Norden (2011,

2012). For example, Lahiri and Wang (2013: 181) describe

resolution as follows: ‘. ..it refers to the ability of a set of

probability forecasts to sort individual outcomes into probabil-

ity groups which differ from the long-run relative frequency.’

Lahiri and Wang (2013: 181) continue by saying that ‘it refers

to an aspect of forecasting skill that discriminates between

individual occasions on which the event of interest will and will

not take place.’ In a loose sense, resolution is about the ‘vari-

ability’ of the forecasts and how that variability matches with

the variation in the realizations.

4Note that a set of forecasts may have resolution power if they

are poorly calibrated (e.g., overly optimistic) or if they are well

calibrated. Resolution is, thus, a dimension of forecasts that is

not part of the conventional definition of overoptimism.

284 A. Hyytinen, J. Lahtonen, and M. Pajarinen

Copyright © 2014 Strategic Management Society Strat. Entrepreneurship J.,8: 283–302 (2014)

DOI: 10.1002/sej