2. Literature review and hypotheses
2.2 Determinants of loss-induced R&D
Prospect incentive
BT has guided much recent research on risky organizational changes,
including firm’s R&D investment decision. This theory proposes that firms’ R&D
7 This survey presents 84% of financial executives agree that the earnings improvement benchmark is important, and there are 69%, 65% agree that analyst forecast and loss avoidance benchmark is important respectively.
13
investments can be attributed to two types of organizational search, i.e., problemistic
search and slack search. However, since my study focus only on loss-induced R&D, I
put emphasis on problemistic search alone8. When firms fail to attain their aspiration
level, they launch problemistic search which aims to solve the performance problems
immediately. Problemistic search induces increased R&D when managers judge that
upgrading firms’ technology and product portfolio can solve the performance
problems, and this judgment happens universally (Greve, 2003). Once firms’
performance are above their aspiration level, they tend to maintain current routines
and have limited motivation to search for anything new (Cyert and March, 1963;
Levinthal and March, 1981).
However, what is the role of risk preference in the schema of loss-induced
R&D? Although BT suggest that performance under aspiration level will increase
managers’ tolerance for risk, i.e., they are willing to take more risk, because they are
under pressure from stakeholders to return to aspiration as soon as possible, the theory
does not indicate whether risk seeking incentive has any association with loss-induced
R&D. To illustrate, I refer to Greve’s (2003) conceptual model for BT, shown in
8 Problemistic search is defined as "search that is stimulated by a problem ... and is directed toward
finding a solution to that problem" (Cyert and March, 1963); whereas slack search refers to search that is triggered by firms’ endowment such as underused financial reserves, capacities, facilities and labor (Levinthal and March, 1981). In my study, slack search (firm’s endowment) may be used as a control variable rather than a primary variable.
14
Figure 1. It shows that R&D is driven by two factors, problemistic search and slack
search as mentioned before, and the latter is neutral from the firm’s performance level
unlike the former shows a direct negative association between firm’s relative
performance to its aspiration level and R&D. With respect to risk preference, denoted
as risk tolerance, I can see that it influences firm’s behavior at the later stage,
“decision making”, but not at the earlier “search” stage, which, in my opinion, implies
that risk preference is irrelevant to R&D decision process and the problem-solving
incentive is the only factor driving loss-induced R&D.
Figure 1: The conceptual model of the behavioral theory. Source: Greve (2003).
Except for the previous question I raised on BT, I have two additional
concerns for the argument of problem-solving incentive. First, to my knowledge, the
15
role of problem-solving incentive in the schema of loss-induced R&D is given and
predetermined, that means researchers following BT usually do not test either the
argument of problem-solving incentive is correct or its quantitative impact on R&D or
future earnings. One of the reasons for this problem could be that problem-solving
incentive lacks theoretical support, or at least such theory has not yet been referred,
and it thus makes problem-solving incentive immeasurable compared with the
incentive derived by firm’s endowment (slack search). Second, the lack of theoretical
support for problem-solving incentive limits the scope of research following BT and
yields limited implications for problem-solving incentive. For example, given the
great interest in the future performance of R&D (either earnings or its variability) in
accounting and economic academy, researchers cannot make any prediction on the
future earnings or its variability of loss-induced R&D since there are no theory
examining the association between such incentive and future performance.
I organize the above problems in BT to a simpler representation: the
overemphasis on problem-solving incentive, which can be problematic, lead to
overlooking on risk seeking incentive when analyzing loss-induced R&D. To address
16
this problem, I first introduce Prospect theory (hereafter PT; Kahneman and Tversky,
1979) and incorporate it into BT9 subsequently.
To illustrate, PT distinguishes two phases in the decision process: an early
phase of editing and a subsequent phase of evaluation. In the first phase, once people
receive different prospects (or gambles, investment projects) to choose, they start to
form a reference point, consisting of the factors from past and present related context
of experience10, and then translate each prospect to a simpler representation. In the
second phase, the edited prospects are evaluated and the prospect of highest value is
chosen. Since the value function used by people to evaluate prospects in phase 2 is the
core argument in PT, I show this function in Figure 2 and provide a further
explanation on its key properties as follows. First, it contains a reference point,
derived from phase 1 and separating the outcome of each prospect into gains and
losses region which is perceived by the individual. Simply speaking, reference point is
9 PT is originally developed to explain individual decision making under uncertainty that cannot be explained by expected utility theory (Von Neumann and Morgenstern, 1944); however, it is well documented that PT also works at firm level (e.g., Fiegenbaum and Thomas, 1988; Chang and Thomas, 1989; Miller and Bromiley, 1990; Jegers, 1991; Sinha, 1994; Johnson, 1994; Gooding et al., 1996).
10 In PT, reference point is the status quo or a value of zero (Kahneman and Tversky, 1979) basically;
however, it is also consisted of multiple factors like aspiration level, and there is no general rule for deciding on such a definition. Most researchers assume a common reference point by measuring industry median or mean of returns over the time period (Fiegenbaum and Thomas, 1988; Fiegenbaum, 1990; Miller and Bromiley, 1990; Jegers, 1991; Sinha, 1994; Johnson, 1994).
17
equal to the aspiration level in BT. Second, the carriers of value are changes in wealth,
rather than final states. That is, I can see a given point on the x-axis as “earning 10
dollars”, which shows a certain performance level. Third, the value function is
concave for gains, convex for losses, being steeper for losses than for gains.
Figure 2: The value function of the prospect theory. Source: Kin Fai Ellick and Jessica Y.Y.
(2005).
The above properties of value function jointly yield a critical implication for
my study, a fourfold pattern of individual’s (firm’s) risk attitudes, derived from
twofold risk attitudes. Specifically, firm is risk seeking (averse) when it perceives that
its performance level is in the losses (gains) region, and its risk attitudes keep varying
among each region, i.e., firm in big losses (small gains) is more risk seeking than that
in small losses (big gains). I first explain the reason of twofold risk attitudes by
showing the risk seeking side. When firm fails to attain its aspiration level (reference
point), it falls in the losses region and the discrepancy (between its performance and
reference point) will persists until firm adapts to this downward shock. It
18
consequently induces two results. First, the perceived losses situation will affect
firm’s way to code the prospect11, which in turn leads firm to adopt riskier prospect.
Second, since the marginal benefit (in terms of individual’s psychological value) at
any given point of the value function always outperforms marginal cost in losses
region, firm has strong incentive to increase their performance by taking riskier
prospect. To further explain fourfold risk attitudes, I refer to the value function’s
convex and concave characteristics. When firm falls in the region of big loss, taking
more risk incurs little, close to zero, marginal losses (in terms of individual’s
psychological value), therefore, firm in big losses is more likely to adopt risky
prospect than firm in small losses. As for firm in big gains region, taking more risk
incurs little, close to zero, marginal gains (in terms of individual’s psychological
value), therefore, firm in big gains is less likely to adopt risky prospect than firm in
small gains.
Supporting fourfold risk attitudes, Greve’s empirical studies (1998, 2003b,
2003c), which is based on US radio industry, Japanese shipbuilding industry and
cross-section data respectively, show a kinked curve which illustrates that the
relationship between firm performance and the probability of strategy (risky) change
has a tipping point at the aspiration level (as shown in Figure 3, from Greve (1998)).
11 To see a complete example, please check Kahneman and Tversky (1979), p.286.
19
This downward sloping curve shows that firms at different performance level do risky
changes, defined as format change, facility and innovation investment respectively,
different in magnitude. Specifically, firms in losses take more risk than those in gains,
and among each region, firms in big losses (small gains) take more risk than those in
small losses (big gains). His proxy for risky change is very similar with R&D, of my
main interest, since R&D is also an investment with high risk (Kothari et al., 2002).
Hence, based on fourfold risk attitudes framework and Greve’s empirical evidence, I
hypothesize that firms with increased risk seeking incentive (in my term, prospect
incentive) invest more in R&D. Such hypothesis thus incorporates PT into BT
automatically since it extends the twofold risk attitudes assumption. My hypothesis is
as follows:
H2. Prospect incentive (based on fourfold risk attitudes framework) drives loss firms
to invest more on R&D.
20
Figure 3: Relation between performance level and probability of change. Source: Greve (1998).
2.3 Future financial performance of loss-induced R&D
Though R&D investments do benefit firms on average according to a variety
of literature no matter from accounting perspective or economic perspective (e.g., Lev
and Sougiannis, 1996; Eberhart et al., 2004; Lev et al., 2006; Long and Ravenscraft,
1993; Vivero, 2002; Branch, 1974; Tassey, 1983; Morbey and Reithner, 1990;
Doukas, 1991; Erickson and Jacobson, 1992; Ito and Pucik, 1993; Johnson and
Pazderka, 1993; Lee and Shim, 1995), when I look specifically at loss firms, situation
seems change. Accounting literature has shown some clues, for example, based on
Hayn (1995), proposing that loss itself could persist for several years, Joos and Plesko
(2004) further shows that persistent loss firms have more R&D. The association
between loss and R&D is tested informally until Li (2011) showing that R&D is
negatively associated with forecast earnings.
Based on my H2, I further argue that prospect incentive may lead loss firms to
invest in unproductive R&D, i.e., it has negative impact on future earnings, according
to PT. Under PT’s assumptions, loss firms will choose investment projects with high
risk (e.g. R&D, Kothari et al., 2002) to bet on the probability of reversing from losses
rapidly. In other words, they care variance in returns more than expected returns.
Research following PT has found that firms with unbalanced risk-return preference
21
presented above, i.e., prospect incentive, perform worse in future. For example,
following Bowman’s finding (1980) which shows a negative association between
organizational risk and future return, researchers (Fiegenbaum and Thomas, 1988;
Chang and Thomas, 1989; Fiegenbaum, 1990; Jegers, 1991; Sinha, 1994; Gooding et
al., 1996; Lehner, 2000) further extend that firms below (above) the reference point
exhibited a negative (positive) relationship between risk and future return.
Accordingly, I infer that prospect incentive on R&D may have negative impact on
future earnings. My hypothesis is as follows:
H3. Prospect (incentive)-based R&D of loss firms has negative impact on future
earnings.