Value effect of rival innovation failure: competition or contagion?

Abstract

This paper investigates how the value of firms is affected by rival R&D projects. In particular, we utilise a unique event (i.e. the failure of the R&D project in Hepatic vaccine in Chongqing Brewery) to investigate the market response to the announcement of failure in R&D investment in a rival firm. The competition (contagion) hypothesis predicts a positive (negative) market reaction. Our results show a strong negative market reaction, indicating that the contagion effect dominates the scene. We also show that the negative effects are more pronounced in firms with core business in the vaccine industry, reinforcing our main evidence supporting the contagion effect. Moreover, our evidence suggests that such an effect is more pronounced in firms with a higher degree of uncertainty. Our study has important contributions to the literature as well as policy implications.

Keywords:

• Contagion
• event studies
• industry competition
• rival R&D

1. Introduction

Innovation is widely viewed as the core competence of a firm. Firms worldwide pour billions into the R&D process. Yet, such investment guarantees no success (Boulding & Morgan, 1997). In fact, most R&D investment ends up in failure (Crawford, 1997). Many have studied the value effect of R&D investment and failure (e.g. Chan, Martin, & Kensinger, 1990; Eberhart, Maxwell, & Siddique, 2004; Hendricks & Singhal, 1997). Little, however, has been devoted to the question as to how the innovation process interacts among competing firms.

Rival firms’ R&D projects have two potential effects on the value of firms. First is the competition effect (Lang & Stulz, 1992). Firms in competitive industries frequently enter into innovation races with many rivals. When a firm successfully completes an R&D project before other firms in the innovation race, these other firms often have to suspend or even abandon similar projects (Chen, Chen, Liang, & Wang, 2013). Such suspension or cancellation of the R&D projects prevents firms from recouping profits from their previous investment, which tends to be irreversible. Therefore, competition can have a substantial impact on R&D-intensive firms. The second effect is the contagion effect (Lang & Stulz, 1992). The news from rival firms may reveal new information on the common technology, or alter the sentiment toward the whole industry. Based on that information, investors adjust their expectation on the likelihood of R&D projects of firms in the same direction as rivals.

The above two effects produce different predictions on the relation between value movements of rival R&D projects in the same industry. While the competition effect predicts a negative association due to the innovation race, the contagion effect indicates a co-movement of value in the same direction. Hence, whether rival firms benefit from other firms’ R&D investment or failure is an empirical question. On the other hand, the competition effect is more likely to dominate the scene among rival firms at the time of successful R&D (e.g. Beath, Katsoulacos, & Ulph, 1989; Gu, 2016).

This paper investigates how the value of firms is affected by rival R&D projects. In particular, we utilise a unique event of the market response to the announcement of failure in a rival’s R&D investment. On 5 December 2011, Chongqing Brewery (Shanghai Stock Exchange: 600132, CB hereafter) halted stock trading. It announced the failure of its R&D on Hepatic B vaccine on 8 December 2011. The stock price of CB fell to less than a quarter of its historic high. It attracted a significant level of investor attention from the capital market. The Baidu Search Index of CB from 2011 to 2016 shows a large spike was observed around the announcement, indicating a significant increase in investor attention on the firm.¹ Therefore, we expect that investors in firms with rival R&D projects are likely to learn from the announcement and adjust their valuation. We empirically test the market reaction of these firms in a short window around the announcement. The results show strong negative market reaction to the news regarding the failure of the clinical trials, indicating that it is the contagion effect that dominates the scene. We also show that the negative value effect is more pronounced in firms whose core business is within the vaccine developing industry. Our evidence also suggests that such an effect is more significant in firms with a greater degree of uncertainty. We document weak evidence that corporate governance plays a role in mitigating the contagion effect.

This paper contributes to the literature in the following ways. First, it adds to the studies on the valuation of R&D investment by introducing the rival-specific risk. Prior studies document a positive relation exists between R&D investment and stock returns (Chen et al., 2013; Hirshleifer, Hsu, & Li, 2013; Lev & Sougiannis, 1996). Gu (2016) argues that the success of an R&D project in a rival firm may increase the likelihood of failure in R&D due to competition, and result in higher ex ante required return on R&D. Zantout and Tsetsekos (1994) and Chen et al. (2013) investigate the value effects on rival firms for increased R&D investment. These two papers both study developed markets yet reach opposite conclusions. This paper, in contrast, proposes and confirms another type of rival risk, i.e. a contagion effect from the failures of rivals, which results in value loss when a rival company fails in similar R&D projects. It is important to emphasise that, ex ante, whether and how a firm’s R&D failure could influence rival firms’ value is an open question. Moreover, the contagion effect here should be differentiated from the ‘spillover effect’ documented in previous studies. For example, Chen et al. (2013) find the firms benefit from R&D investment made by other firms in profitability and long-run stock performance. Griliches (1979) and Bernstein and Nadiri (1988) document that costs of rival firms decline as result of R&D spillover. Jaffe (1986) shows that when the potential R&D spillover pool increases by 1%, the profit of other firms increases by 0.3%. The contagion effect is different from the spillover effect in that it is caused by the failure of a rival’s R&D and introduces negative externality, while the spillover effect resulting from technological success produces positive externality. Moreover, the contagion effect could be explained with either rational reasons, for example, the failure of technology, or through irrational sentiments, while the spillover effect is only explained by rational reasons such as the development of technology, as in previous literature.

Second, we contribute to studies on interactions between the strategy of R&D investment and industry competition. Managers make decisions on R&D investment not only based on the cost and benefit for the company itself but also the expected interactions with industry peers. While most studies in this stream develop theoretical models (e.g. Aghion, Bloom, Blundell, Griffith, & Howitt, 2005; Lin & Zhou, 2013), empirical evidence is limited. Among this evidence, McGrath and Nerkar (2004) document a curvilinear relation between the extent of competition and R&D investment from the perspective of real options. Our study provides new evidence by utilising a unique setting of the failure of a rival’s R&D project to differentiate the effects from competition and contagion.

Our study also fits in a broader literature of intra-industry information transfer. Previous studies document a significant value effect on the industry peers when the information of a firm’s risk and value is revealed in announcements of earnings (Foster, 1981), merger bids (Cai, Song, & Walkling, 2011), bankruptcy (Lang & Stulz, 1992) and a new stock offering (Szewczyk, 1992), and in the spread of credit default swaps (Jorion & Zhang, 2007). These studies show that the contagion and competition effects coexist. Our study extends this stream of literature by adding strong evidence of the contagion effect in the setting of R&D failure.

Last but not the least, the current study adds to the literature on R&D in the Chinese capital market. Innovation in China has made great progress in recent years (see a detailed review in Section 2). However, research on innovation in China is very scarce and mainly focuses on institutional factors such as government subsidy (Guo, Guo, & Jiang, 2016) and property rights (Fan, Gillan, & Yu, 2013). The current study adds to this limited yet burgeoning literature.

In addition to its academic contribution, our study also carries import implication to investors, firms, and regulators. The contagion effect that we have documented gives a warning against putting all the eggs in one basket to investors keen on innovative firms. An innovative firm, on the other hand, may want to release its R&D information strategically, taking the potential market reaction to both itself and its rivals into consideration. The contagious nature of R&D news suggests that policy-makers may offer a new policy to prevent a catastrophic market crash in the face of negative R&D news from one firm.

The rest of the paper is organised as follows. Section 2 introduces the background. Section 3 reviews the literature and develops our hypotheses. Section 4 introduces our sample and empirical design. Section 5 presents our main empirical results. Section 6 reports further analysis and robustness tests. Section 7 concludes.

2. Background

According to ‘The report on status and development of PCT system in China (Year 2014)’ released by State Intellectual Property Office (SIPO) of China, China is already among the leading countries filing patent applications through the Patent Cooperation Treaty (PCT), a globally accepted alliance for patent filing, applying and protecting. According to the same report, the number of PCT applications from China increased to over 25,000, earning it the third spot, after the US and Japan, in 2014. According to the State Intellectual Property Office (SIPO), the target in 2015 was:

the annual quantity of applications for patents for inventions, utility models and designs will reach 2 million. China will rank among the top two in the world in terms of the annual number of patents for innovations granted to the domestic applicants, and the quality of patents filed will further improve.

All these factors show that China is making remarkable progress in the intellectual property arena, although a large gap with developed markets remains, which encourages more investment in R&D from the private sectors.

In China, few pharmaceutical research projects have triggered a frenzy as large as that for hepatitis B vaccines in the capital market (Chen & Xiao, 2006). Hepatitis B is still one of the most serious health issues China faces. Statistics show that spending on hepatitis B treatments equals about 1% of the country’s GDP every year (Xie, 2006).

According to Chen and Xiao (2006), existing treatments, including antiviral drugs and injections of interferon, only limit the ability of the virus to replicate and cause infections in the liver, but they cannot rid patients of the virus itself. Treatments for chronic hepatitis are protracted and expensive. As soon as treatment stops, the virus residing in the body can cause infections to recur, and the only hepatitis B vaccines in existence are prophylactic, i.e. they only prevent infection (Wang, He, Wang, & Chen, 2014).

A therapeutic vaccine for hepatitis B is not available. Hence, it is not surprising that Chinese firms pour lots of cash into searching for a cure. Such firms include CB, which is from a seemingly unrelated industry. In 1998, one year after CB’s IPO in the capital market, the firm established a subsidiary named Jia Cheng Co. to initiate research into a vaccine for hepatitis B. Although CB itself is not a firm in the biopharmaceutical industry, its investment in the Hepatic vaccine project is apparently an important strategy of this company. As of the year end in 2011, total investment implementing this R&D project was up to about 23 million yuan, funded in the form of equity and loans, and was worth over 6.3% of total assets of CB. CB also announced that it had invested 31 million yuan for the period-II clinic test in year 2010 alone, which is about 1.8% of the total revenue.

However, in December 2011, the team’s second phase of clinical trials showed that, apart from one minor indicator, the differences in the results of patients who received the injection and those who received a placebo were not statistically significant. After the results were released on 19 December 2011, CB’s stock fell by the daily maximum for nine straight trading days and the market value of 25,000 million yuan disappeared (see Figure 1). The company warned that its vaccine research project faced major risks including the possibility of not proceeding to the third phase of clinical trials. In May 2012, the management of CB formally announced that the project would be closed permanently due to the failure of clinical trials.

Figure 1. Price of Chongqing Brewery (Source: created by the authors from published data).

3. Related literature and hypothesis development

Prior research investigates the relation between R&D investment and firm value, and generally agrees on a positive correlation between the two (e.g. Grabowski & Mueller, 1978; Hana & Manry, 2004; Hirschey & Weygand, 1985; Lev & Sougiannis, 1996; Pakes, 1985). The slump in CB’s price clearly confirms the notion that the Chinese capital market values genuine innovation. However, in addition to the benefit and risk of the project itself, the value of R&D projects is also affected by the progress of rival projects. The prior empirical literature has studied the rival value effects on the announcements of important corporate information such as bank bankruptcy (Aharony & Swary, 1983; Akhigbe & Madura, 2001; Ferris, Jayaraman, & Makhija, 1997; Gay, Timme, & Yung, 1991; Jorion & Zhang, 2007; Lang & Stulz, 1992), omitting dividends (Impson, 2000), privatisation (Chen, Li, & Moshirian, 2005; Otchere, 2005; Otchere & Chan, 2003), and downgrading of corporate bonds (Akhigbe, Madura, & Whyte, 1997). These studies suggest the existence of a competition effect or a contagion effect, or both, in the relationship among the competitors. It is not clear how the announcement of the failure in the R&D project at CB affected its rival companies. We conjecture that there are at least three possibilities.

3.1. Competition effect

Investment in R&D is one of the most important activities driving companies’ long-term viability. Firms in competitive industries frequently enter into innovation races with many rivals. Beath et al. (1989) argue that competitors’ threat is one of the critical determinants of a firm’s R&D expenditures. While a firm might myopically overinvest in reaction to the competitive threat, an R&D investment announcement may be taken as a signal that the announcing firm is moving ahead in the race to be the first to innovate. Accordingly, the competitive structure of the industry may be altered to its advantage.

When one firm successfully completes an R&D project before other firms in an innovation race, these other firms often suspend or even abandon similar projects (Chen, Huang, & Wang, 2014). Suspending or abandoning an R&D project significantly reduces firm value, for doing so prevents projected cash flows associated with the R&D project from being realised, and initial investment tends to be irreversible (Chen et al., 2013). Therefore, competition can have a substantial impact on R&D-intensive firms (Gu, 2016). Empirical evidence on the relationship between innovation and competition is limited and produces mixed results (Aghion et al., 2005; Boldrin, Allamand, Levine, & Ornaghi, 2011; Correa, 2012; Correa & Ornaghi, 2014; Hashmi, 2013). From the perspective of a competition effect, it is possible that investors view CB’s failure in the R&D project as providing opportunities for its rivals. In fact, it is typical for the first achiever to seize the lion’s share of the premium in the market for new medicine research (Chen et al., 2014). Hence, CB’s failure in the R&D project of clinical trials and departure from this market opens a door for the rest. Moreover, it is possible that rival companies build their own research on CB’s failure in the R&D clinical trials project and jump ahead. Thus, we establish the following hypothesis:

H1a: CB’s announcement of the failure of clinical trials has a positive impact on the value of rival firms. (Competition hypothesis.)

3.2. Contagion effect

Rival firms may also suffer a loss in their stock price as CB does. That is, the bad news of a firm may result in a contagion effect on its industry peers. For example, Lang and Stulz (1992) demonstrate that intra-industry rival stock price reactions to a competitor’s bankruptcy are significantly negative on average. Hertzel and Officer (2012) find the spreads on new and renegotiated corporate loans are significantly higher when the loan originates in the two years surrounding bankruptcy filings by industry rivals. Jorion and Zhang (2007, 2009) expand the investigation by examining the effect of financial distress on a credit default swap (CDS) spreading to industry competitors and suppliers.

The contagion effect may go through two channels in the context of R&D failure. The first channel is the sentiment of investors. Baker, Wurgler, and Yuan (2012) document that the contagion effect of investors’ sentiment would transfer across countries. Similarly, the pessimistic emotion aroused from the failure of R&D projects could spread to other firms with similar investments. Second, the failure of an R&D project may reveal the technical difficulty or unobservable cost of a specific proposal. Since the bio-pharmaceutical industry is an innovation-intensive industry that is characterised by high risk and a long R&D cycle, it is hard to predict all costs and technical difficulties fully in the early stages. For example, the development of a new medicine must go through four stages (pre-clinical, clinical test, trial production, and mass production), and the duration from new medicine proposals to acquiring approval for production will take at least three to six years or over ten years in China. Thus, innovation in the bio-pharmaceutical industry involves great uncertainty and the assessment on risk and cost is time-dependent. Dimasi, Grabowski, and Hansen (2016) document that the likelihood of a drug being eventually approved is only 11.83%. Therefore, CB’s failure in its vaccine project is likely to lower the investors’ expectation of success for the same technology applied in rival firms. This, in turn, lowers the expected value of future cash flow and hence the stock market value of the firm. Thus, we establish the following hypothesis based on the above argument:

H1b: CB’s announcement of the failure of clinical trials has a negative impact on the value of rival firms. (Contagion hypothesis.)

Finally, it is also likely that the announcement simply has no impact on the rival firms, if both the competition effect and contagion effect take place and cancel each other out.

3.3. The role of uncertainty

While not explicitly stated, the role of uncertainty is crucial in our discussion above. In fact, both the competition hypothesis and the contagion hypothesis are built on the assumption that the market cannot perfectly predict the outcome of the rival companies’ own innovation process. The capital market would not update its belief on the success of these innovation processes if there is no uncertainty in the R&D process in the first place. The larger uncertainty firms have, the higher the demand on the information of rivals from the investors. Hence, we predict that the effect from CB’s failure in the R&D project of clinical trials will be more pronounced for the rival companies with more uncertainty. We formally state our hypothesis below.

H2: The impact of CB’s announcement of failure in the R&D clinical trials project on the value of rival firms is more pronounced for firms with a higher degree of uncertainty.

4. Sample and empirical design

4.1. Sample

To test our hypothesis, it is crucial to define precisely the scope of rival firms. We select the firms in the bio-medical industry which are most likely to compete with CB in developing the vaccine. We rely on the industry codes of Shenyin Wanguo to filter out 164 firms in the bio-medical industry that are listed before CB’s announcement of the failure of its R&D project. We exclude firms whose core businesses are medical tools, medical business, or medical service as they do not directly compete with CB in the medicine industry. After such filtering, we are left with 136 sample firms, among which, 53 are from the chemical medicine industry, 29 from the bio product industry, and the remaining 54 are from the traditional Chinese medicine (TCM) industry.

4.2. Empirical design

We employ the event study method to investigate the change of firm value. CB stopped stock trading on 27 November 2011, and announced a delay in the disclosure on results of clinical tests on 5 December 2011, which delivered a signal of high uncertainty to the market. On 8 December 2011, the failure of the clinical trials was finally announced and the trading restarted. Thus, we set the period from 5 December 2011 to 8 December 2011 as the event ‘period’. We use the market adjusted cumulative abnormal return (CAR)² during the (–1,1) of the event period, that is, the window period is from the first trading day before 5 December 2011 to the first trading day after 8 December 2011.

With the CAR, we employ two empirical designs to test our hypotheses. First, we test whether the CARs for our sample firms are on average positive or negative. Our competition (contagion) hypothesis predicts a positive (negative) average return for these rival firms. Second, we test whether the effect is more pronounced for firms whose core business more closely resembles the nature of the CB’s R&D project. We create a dummy variable, Vaccine, that indicates whether a firm also develops vaccine. We manually search annual reports, in the year prior to the announcement, for the keyword ‘vaccine’ and code the variable Vaccine to be 1 if we find at least one occurrence of vaccine in a (rival) firm’s R&D projects and 0 otherwise.

We then estimate the following model:(1) $$CAR=Vaccine+Controls$$(1)

where controls include firm age, size, ROA, and leverage in the year prior to the announcement following the prior literature (e.g. Bradley & Yuan, 2013; Gleason, Thorne, & Bruce, 2008). All the variables are defined in Table 1. We expect a positive (negative) and significant coefficient on Vaccine if the competition (contagion) hypothesis dominates.

Table 1. Definition of variables.

Variable	Definition
CAR	Cumulative abnormal return, calculated by cumulating stock returns minus returns of the market index on the listing stock exchange in (–1, 1) event window. The event days are 2011.12.5-2011.12.8
Vaccine	A dummy variable equal to one if a firm’s R&D involves vaccine
Vaccine_Percent	The ratio of the investment in R&D projects on vaccine and the total investment in R&D projects in process
GBYM	A dummy variable equal to one if a firm’s R&D involves hepatic vaccine
QTYM	A dummy variable equal to one if a firm’s R&D involves non-hepatic vaccine
SWZP	A dummy variable equal to one if a firm’s R&D involves non-vaccine medicine
Log_Board	The natural logarithm of the size of board
Nwage	The proportion of board members that do not receive pay from the firm
Analyst recommendation	The mean of analyst recommendations in the 6 months before the announcement, equal to 5 if the recommendation is ‘strong buy’, equal to 4 if the recommendation is ‘buy’, equal to 3 if the recommendation is ‘neutral’, equal to 2 if the recommendation is ‘sell’, equal to 1 if the recommendation is ‘strong sell’
Analyst forecast dispersion	The dispersion of three-year growth rate of earnings per share forecast by analyst
Age	Number of years since a company’s initial public offering
ROA	Return on assets, measured as operating income scaled by total assets
Size	Company size, measured as the natural logarithm of total assets
Leverage	Financial leverage, measured as total liabilities, scaled by total assets

To test our Hypothesis 2, we augment equation (1(1) $$CAR=Vaccine+Controls$$(1) ) with proxies for uncertainty of a firm’s innovation and, more importantly, their interactions with the Vaccine dummy. That is, (2) $$CAR=Vaccine+Uncertainty+Uncertainty\ast Vaccine+Controls$$(2)

In Equation (2(2) $$CAR=Vaccine+Uncertainty+Uncertainty\ast Vaccine+Controls$$(2) ), our variable of interest is the interaction term between uncertainty and the Vaccine dummy. Our Hypothesis 2 predicts that the sign on this interaction term should be positive (negative) if the competition (contagion) effect dominates. Following the prior literature, we proxy a firm’s future economic uncertainty with two analyst related measures (e.g. Barron & Stuerke, 1998). It is worth pointing out that by using the three measures, we are assuming that a larger portion, if not all, of a firm’s future economic uncertainty comes from R&D. The first measure is the dispersion of forecast three-year growth rate of earnings per share (Analyst forecast dispersion). Analyst forecast dispersion is shown to be indicator of a firm’s future economic uncertainty (Barron & Stuerke, 1998). We deviate from the prior literature by focusing on a longer horizon since R&D takes a significant time before being realised. The second measure is the mean of the analysts’ recommendations (Analyst recommendation) in the 6 months before the announcement. This measure indicates analysts’ confidence in the value of the firm. Hence, a higher level of recommendation indicates lower uncertainty.

5. Description of data and empirical results

5.1. Description of data

For the list of 136 rival firms, we obtain daily trading as well as the accounting data necessary to construct the variables described in Table 1 from the China Stock Market and Accounting Research database (CSMAR).

Figure 2 shows that CARs of sample firms decline averagely around the event window of (–1, 1). Panel A of Table 2 shows that the mean and median CAR for the 136 sample firms are –2.33% and –2.69%, respectively. Both the mean and median CARs are negative and statistically significant. This piece of evidence lends support to the contagion hypothesis. That is, the market punishes the rival firms when CB announces the failure of its R&D project. Given that the average market value of the 136 firms is 6.8 billion RMB, the firms on average lose more than 0.15 billion RMB in market value during the short window around the announcement of CB’s failure in the R&D project. Hence, the losses these firms suffer are economically meaningful in addition to its statistical significance.

Figure 2. Cumulative abnormal return of rival firms around Chongqing Brewery’s announcement of failure in clinical trials.

Table 2. Summary statistics.

Variables	N	Mean	Std	Min	P1	P25	Median	P75	P99	Max
Panel A: Summary statistics of CAR
CAR	136	–2.33%***	4.25%	–11.67%	–9.93%	–5.56%	–2.69%***	–0.07%	7.13%	11.38%
		(–6.40)					(–7.40)
Panel B: Summary statistics of other variables
Vaccine	136	0.20	0.40	0.00	0.00	0.00	0.00	0.00	1.00	1.00
Vaccine _Percent	136	0.03	0.16	0.00	0.00	0.00	0.00	0.00	1.00	1.00
Age	136	7.90	6.10	0.00	0.00	1.00	7.50	13.00	19.00	20.00
ROA	136	0.10	0.09	–0.21	–0.17	0.05	0.09	0.13	0.30	0.40
Size	136	21.13	0.96	19.03	19.07	20.61	21.09	21.73	23.29	23.55
Leverage	136	0.32	0.20	0.01	0.02	0.16	0.30	0.44	0.87	1.17
GBYM	136	0.07	0.26	0.00	0.00	0.00	0.00	0.00	1.00	1.00
QTYM	136	0.13	0.33	0.00	0.00	0.00	0.00	0.00	1.00	1.00
SWZP	136	0.43	0.50	0.00	0.00	0.00	0.00	1.00	1.00	1.00
Board	136	9.37	1.87	5.00	5.00	9.00	9.00	10.50	15.00	15.00
Nwage	136	0.14	0.13	0.00	0.00	0.00	0.11	0.23	0.43	0.44
Analyst recommendation	111	1.02	0.44	0.00	0.00	0.88	1.00	1.29	2.00	2.00
Analyst forecast dispersion	101	0.09	0.08	0.00	0.01	0.04	0.06	0.10	0.37	0.64

Notes: This table presents summary statistics of cumulative abnormal return (CAR) of rival firms around CB’s announcement of R&D failure (Panel A) and other variables (Panel B). Definitions of variables are in Table 1. t-statistics are given in brackets with * denotes significant at the 10% level, ** denotes significant at the 5% level and *** denotes significant at the 1% level.

Panel B of Table 2 presents the summary statistics for the variable used in our main analysis. The mean of the Vaccine dummy is 0.20, which indicates that 20% of our sample firms develop vaccines. The mean and median of firm age are 7.9 and 7.5, respectively. Our sample firms are, on average, profitable, with a mean ROA of 0.10. The average leverage of the sample firms is about 34%. On the whole, all the variables show little skewness but large variation.

Table 3 presents the correlation among our key variables. There are two points worth mentioning. First, the correlation between CAR and Vaccine is –0.15 and statistically significant at the 10% confidence level. This lends preliminary support to our contagion effects. Second, none of the correlations are greater than 0.7, leaving our regression analysis with little chance of multicollinearity issue.

Table 3. Correlation matrix.

	CAR	Vaccine	Vaccine_Percent	GBYM	QTYM	SWZP	Log_Board	Nwage	Analyst forecast dispersion	Analyst recommendation	Age	ROA	Size	Leverage
CAR	1.000	–0.158*	–0.202**	–0.099	–0.112	–0.025	0.059	0.015	0.016	0.068	–0.037	0.164*	–0.079	–0.042
Vaccine		1.000	0.368***	0.566***	0.759***	–0.436***	–0.049	–0.053	–0.081	–0.064	0.036	0.040	0.198**	0.048
Vaccine_Percent			1.000	0.293***	0.213**	–0.160*	–0.022	–0.042	0.241	–0.067	–0.143*	0.028	0.083	–0.139
GBYM				1.000	–0.106	–0.247***	0.016	–0.162*	–0.101	–0.130	0.005	0.134	0.136	–0.075
QTYM					1.000	–0.331***	–0.071	0.064	–0.016	0.022	0.039	–0.058	0.131	0.117
SWZP						1.000	–0.052	–0.048	0.123	–0.168*	–0.068	0.100	–0.129	0.056
Log_Board							1.000	0.157*	–0.130	–0.069	–0.185**	0.243***	–0.080	0.045
Nwage								1.000	–0.096	–0.024	0.122	–0.174**	–0.040	0.060
Analyst forecast dispersion									1.000	–0.149	0.070	–0.400***	–0.144	0.094
Analyst recommendation										1.000	0.040	0.095	0.239**	0.050
Age											1.000	–0.375***	0.317***	0.466***
ROA												1.000	–0.034	–0.396***
Size													1.000	0.076
Leverage														1.000

Notes: This table presents correlation matrix of all variables. Definitions of variables are in Table 1. * denotes significant at the 10% level, ** denotes significant at the 5% level and *** denotes significant at the 1% level.

5.2. Main results

The OLS estimation results of Equation (1(1) $$CAR=Vaccine+Controls$$(1) ) are reported in Table 4. The coefficient on Vaccine is negative and statistically significant at the 10% confidence level. Given the small size of our sample, the statistical significance of our tests is relatively strong. Consistent with the results in Table 2, the regression results lend support to the contagion hypothesis. It is worth pointing out, however, that our results do not necessarily rule out the competition hypothesis. In fact, both competition effect and contagion effect can take place. We will observe a negative value effect as far as the contagion effect dominates.

Table 4. Regressing CAR on Vaccine dummy.

	CAR
	Coefficient	t-value
Intercept	0.021	0.24
Vaccine	–0.017*	–1.80
Age	0.000	0.46
ROA	0.098**	2.06
Size	–0.002	–0.63
Leverage	0.005	0.26
Adjusted R^2	2.30%
F-test	0.16
N	136

Note: This table reports the results of model (1), and the definitions of variables are in Table 1. * denotes significant at the 10% level, **denotes significant at the 5% level and ***denotes significant at the 1% level.

To test our Hypothesis 2, we estimate the model equation (2(2) $$CAR=Vaccine+Uncertainty+Uncertainty\ast Vaccine+Controls$$(2) ), which is essentially an augmented model of Equation (1(1) $$CAR=Vaccine+Controls$$(1) ). As discussed in the hypothesis development section, the contagion effect may take effect if investors lower their expectation on the success rate or experience negative sentiment about vaccine research. Either way, the contagion effect should be more pronounced in the firms with a higher degree of uncertainty. First, in column 2 of Table 5, we find that the coefficient on interaction between dispersion of forecast long-term growth of earnings per share and Vaccine is significantly negative, which shows the value declines more when uncertainty is high. Then, in column 3 of Table 5, we investigate whether a positive analyst recommendation reducing uncertainty can mitigate the decline in value. In particular, the coefficient on the interaction term between analyst recommendation and the Vaccine dummy is positive and statistically significant. That is, the contagion effect is smaller for a firm with previous positive analyst recommendations.

Table 5. The role of uncertainty.

	CAR
	(1) Analyst forecast dispersion	(2) Analyst recommendation
Intercept	–0.098	–0.027
	(–0.79)	(–0.25)
Vaccine	0.012	–0.062***
	(0.75)	(–2.79)
Vaccine*Uncertainty	–0.434**	0.043**
	(–2.27)	(2.11)
Uncertainty	1.169	–0.008
	(2.20)	(–0.78)
Age	0.000	0.001
	(0.47)	(0.89)
ROA	0.169**	0.136***
	(2.14)	(2.76)
Size	0.002	–0.000
	(0.44)	(–0.01)
Leverage	–0.020	–0.016
	(–0.75)	(–0.61)
Adjusted R^2	7.33%	7.90%
F-test	0.05	0.03
N	101	111

Notes: This table reports results of model (2), and uncertainty is measured with Analyst forecast dispersion and Analyst recommendation. The definitions of variables are in Table 1. t-statistics are given in brackets with *denotes significant at the 10% level, **denotes significant at the 5% level and ***denotes significant at the 1% level.

6. Further analysis and robustness

6.1. Further analysis on corporate governance

The previous results are consistent with the predictions of the contagion hypothesis. In this section, we investigate whether corporate governance plays a role in the contagion effect. Appropriate corporate governance mechanisms reduce agency costs and discipline managers’ behaviour. Such mechanisms also enhance the quality of R&D investment and promise better innovative outputs ceteris paribus (e.g. Baysinger, Kosnik, & Turk, 1991; Francis & Smith, 1995; Hill & Snell, 1988; Holmstrom, 1989). Hence, it is likely that firms with weak corporate governance systems suffer more negative contagion from the news of CB. We empirically test this conjecture by introducing interaction terms of the Vaccine dummy and two (weak) corporate governance measures. The two corporate governance measures are the size of the board and the ratio of board members that do not receive pay from the firm. A larger board is documented to be associated with a higher degree of agency problems (Coughlan & Schmidt, 1985; Jensen, 1993; Murphy, 1985; Yermack, 1996). Board members that do not receive pay from the firm are usually nominated by the controlling shareholders. As such, they are less likely to monitor the controlling shareholders. Hence, we expect the coefficients on the interaction terms to be negative. The empirical results reported in Table 6 confirms our conjecture.

Table 6. The role of corporate governance.

	CAR
	(1) Log_Board	(2) Nwage
Intercept	–0.021	–0.010
	(–0.22)	(–0.12)
Vaccine	0.345***	0.005
	(2.68)	(0.43)
Vaccine*Weak Governance	–0.164***	–0.182**
	(–2.81)	(–2.40)
Weak Governance	0.021	0.045
	(1.04)	(1.39)
Age	0.000	0.000
	(0.31)	(0.30)
ROA	0.075	0.109**
	(1.56)	(2.30)
Size	–0.003	–0.001
	(–0.70)	(–0.39)
Leverage	0.009	0.016
	(0.41)	(0.76)
Adjusted R^2	6.50%	5.10%
F-test	0.03	0.06
N	136	136

Note: This table reports results of the following OLS regressions:

CAR = Vaccine + Vaccine*Weak Governance + Weak Governance + Controls, Weak Governance is weak corporate governance measured by Log_Board and Nwage. All the definitions of variables are in Table 1. t-statistics are given in brackets with *denotes significant at the 10% level, **denotes significant at the 5% level and ***denotes significant at the 1% level.

We also use alternative measures to represent corporate governance, such as auditor reputation, percentage of independent directors, duality of CEO and chairman, managerial shareholdings and composite score on corporate governance as in Haß, Vergauwe, and Zhang (2014), to substitute the above two corporate governance measures (the size of the board and the ratio of board members that do not receive pay from the firm) in additional tests. However, none of them is found to have a significant impact on market reactions to CB’s failure in the R&D project, indicating that the foregoing conclusions are sensitive to measures of corporate governance. For parsimonious reasons, we do not report the results of additional tests here.

6.2. Exploration of effect from success of rival R&D

Although the main topic of this paper is the effect from failure of a rival R&D, we also attempt to examine the effect of success of a rival R&D for comparison. However, it is difficult to identify an exact point of time when an innovation process is successful. One may argue that the time of successful patent approval is the time of success, while others may suggest the time of new product marketisation or even the time of realising economic benefit from the innovation.

Nevertheless, we attempt to empirically test the scene among rival firms at the time of successful R&D. As discussed, it is difficult to identify a time of success. So we take one step back by focusing on a series of good news announcements by CB (e.g. progress of the CB’s R&D projects) before the final failure is revealed. We use the same event study design in our investigation. Table 7 reports the summary statistics of the market reactions to the rivals around these announcements. The cumulative abnormal returns (CAR) of the rivals for a window of (–5, 5) range from –0.539% to 0.336%, none of which are statistically different from 0. These mixed results can be driven by either the lack of the competition effect or the impreciseness in identifying good news (or both).

Table 7. Market reactions of rival companies to the progress of CB’s R&D project.

Event	Passed the test of the Bio-Med Bureau	Initiation of clinical test	Completion of Phase I clinical test
Date of announcement	Jan19, 2002	Nov 28, 2003	Jun28, 2005
Mean of CAR for rivals	–0.54%	0.34%	–0.37%
T value	–0.53	0.54	–0.47
# of Rivals	62	66	77

Note: This table presents market reactions on rival companies to the progress of Chongqing Brewery’s R&D project, including passing the test of the Bio-Med Bureau (19 January 2002), initiation of clinical test (28 November 2003) and completion of Phase I clinical test (28 June 2005). Event window is (–5, 5).

6.3. Robust tests with alternative measures on vaccine

We further use alternative measures on vaccine to check the robustness. First, we categorise firms into more groups based on how close a firm’s R&D is to that of CB’s Hepatic vaccine. In our main analysis, we create one dummy variable, Vaccine, which indicates whether the firm of concern is in search of vaccines. In this section, we create three dummy variables. The first is a dummy variable that takes the value of 1 if a firm’s R&D involves hepatic vaccine. The second is a dummy variable that takes the value of 1 if a firm’s R&D involves non-hepatic vaccine. The third is a dummy variable that takes the value of 1 if a firm’s R&D involves non-vaccine medicines. We replace the vaccine dummy with the three dummies in equation (1(1) $$CAR=Vaccine+Controls$$(1) ). Arguably, the more a firm’s R&D resembles CB’s hepatic vaccine, the more likely the firm suffers the contagion effect. The estimation results reported in column (1) of Table 8 confirm our conjecture. Only the two vaccine dummies enter the regression with negative coefficient statistically significant. Moreover, the (absolute) magnitude of the Hepatic vaccine dummy is the largest.

Table 8. Alternative measures of Vaccine.

	CAR
	Column 1	Column2
Intercept	0.028	0.028
	(0.34)	(0.34)
GBYM	–0.028*
	(–1.89)
QTYM	–0.021*
	(–1.79)
SWZP	–0.013
	(–1.60)
Vaccine_Percent		–0.052**
		(–2.30)
Age	0.000	0.000
	(0.39)	(0.21)
ROA	0.112**	0.086*
	(2.32)	(1.82)
Size	–0.002	–0.003
	(–0.68)	(–0.70)
Leverage	0.010	–0.001
	(0.47)	(–0.06)
Adjusted R^2	2.80%	3.74%
F-test	0.16	0.08
N	136	136

Notes: This table reports results of the following OLS regression:

CAR = Alternative Measures on Vaccine + Controls, and the definitions of variables are in Table 1. *Denotes significant at the 10% level, **denotes significant at the 5% level and ***denotes significant at the 1% level.

Second, the dummy variable of Vaccine used in the main regressions does not reflect the relative importance of the vaccine project in the sample firms. Thus, we hand-collected the amount of investment on R&D projects of vaccine and the total investment on all the R&D projects in progress. We use the ratio of these two to measure the stake the company has on the vaccine project as an alternative measure of Vaccine. The result is consistent with the dummy variable of vaccine, showing a significant negative sign. The result is reported in column (2) Table 8.

7. Conclusion

This paper investigates how the value of firms is affected by a rival R&D project. In particular, we utilise the market response to the announcement of the failure in R&D investment on the hepatic vaccine in CB. Our results show a strong negative market reaction to the failure of a project in a competitor, indicating that it is the contagion effect that dominates the scene. We also show that the negative value effects are more pronounced in firms with a higher degree of uncertainty and whose core business is within the vaccine development industry, reinforcing our main evidence supporting the contagion effect. Moreover, our evidence suggests that such an effect is more pronounced in firms with weaker corporate governance (but is sensitive to different governance mechanisms).

Nevertheless, our study has its limitations. First, our results have been based on one single event and one single industry. Second, our results are based on one very extreme scenario, i.e. the entire failure of a R&D project. As such, one needs caution when generalising our results to another industry or in cases where the news is not as severe as in our sample. The current study leaves to future research the task of verifying the generalisability of our results.

Acknowledgements

Jianqiao Hong, Haoping Xu, and Xin Zhang acknowledge financial support from the National Natural Science Foundation of China (grant numbers: 70972043, 71272073, and 71572047, respectively).

Notes

* Paper accepted by Kangtao Ye.

¹ http://index.baidu.com/. A snapshot of the search is available upon request from the authors.

² Table 1 provides the details of the process to obtain CAR.