Crisis-induced innovation and crisis-induced innovators

ABSTRACT The literature on the persistence of innovation focuses on whether firms reduce, increase, or maintain their innovation activity over time, and in particular through cyclical downturns and crisis periods. What this literature tends to underemphasise is that innovation is also a tool firms can actively use to manage the unforeseen circumstances that arise in times of crisis. We argue and find that this kind of crisis-induced innovation displays patterns that are both similar to, and different from, the innovation behaviours found in more stable periods. More firms turn to innovation, but with important differences in intensity and expected long-run returns. Pre-crisis experience in innovation and organisational agility are key characteristics leading to higher levels of innovation in response to a crisis, higher likelihood of increasing the resources allocated to innovation during a crisis, and higher expected post-crisis value from the innovations undertaken in response to a crisis.


Introduction
At least since Schumpeter (1911;1942) considerable scholarly interest has been devoted to how economic crises and recessions affect innovation (Amore 2015;Archibugi, Filippetti, andFrenz 2013a, 2013b;Barlevy 2004;Geroski, Van Reenen, and Walters 1997).Key questions in this literature have been to determine whether aggregate innovation investments and output are pro-or countercyclical (Archibugi, Filippetti, andFrenz 2013a, 2013b;Caballero and Hammour 2001;Cincera et al. 2012;Filippetti and Archibugi 2011;Guellec and Wunsch-Vincent 2009;Laperche,1911 Lefebvre, andLanglet 2011;Paunov 2012), which firm and contextual characteristics make a firm more likely to be a persistent or countercyclical innovator (Antonioli and Montresor 2021;Cefis 2003;Cincera et al. 2012;Spescha and Woerter 2019), the rewards attributable to those that are persistent innovators (Madrid-Guijarro, García-Pérez-de-Lema, and van Auken 2013), and the impact of recessions and crises on firms' post-crisis innovative performance, behaviour, and capabilities (Paunov 2012;Zouaghi, Sánchez, and Martínez 2018).Underlying the better part of this literature is the notion that faced with a downturn some firms decide to cut innovation activities and use available resources elsewhere -including to ensure survival -while some firms are willing and able to press on with their innovation agenda -or even to speed it up.
What tends to be ignored in the innovation persistence literature is that innovation is also a tool managers can use to handle the unforeseen consequences of a sudden change in business conditions (Wenzel, Stanske, and Lieberman 2020).We argue that an innovation response to an economic downturn does not necessarily imply maintaining or accelerating an already existing innovation agenda (although occasionally it might).Rather, it will often involve ad hoc and unplanned innovation initiatives -unforeseen before the crisis -motivated by the specific changes in business conditions a given firm is facing, coupled with its inventory of ordinary and dynamic capabilities.Such crisis-induced innovation responses will in many cases have a shorter planning-, development-, and implementation period than innovations that takes place in ordinary times, simply because they are aimed at solving urgent and unexpected challenges or opportunities created by the crisis.While innovation persistence is usually explained in terms of various forms of path dependency (Crespi and Scellato 2014), we argue that using innovation to manage unexpected crises will often also require some measure of path independency and organisational agility.
The capacity to undertake crisis-induced innovation can potentially provide managers with an extra degree of freedom with respect to the responses at their disposal, but this capacity is likely to be heterogeneously distributed across firms.In this paper, we therefore examine the extent to which firms use innovation as they navigate the impact of the COVID-19 recession.In doing so, we aim to shed light on an aspect of innovation behaviour that the literature at the intersection of business cycles and innovation has largely left unexplored.Our main interest is to identify key firm-level characteristics that increase the use of innovation as a crisis management tool.
We examine the impact of such firm-level characteristics on the level of crisis-induced innovation undertaken, the expected long-term value of those innovations, and expected changes in innovation investments.
Our primary data source is a survey of 872 Norwegian CEOs, capturing their responses to the COVID-19 recession, and key firm characteristics before the crisis hit.
Our key findings are that a recession incentivise firms to innovate.Firms with a prerecession strategy emphasising innovation as well as firms with higher levels of organisational agility are significantly more likely to introduce crisis-induced innovation, across all innovation types.Not only are these firms more likely to innovate in response to the crisis, we also find that the expected post-COVID value of those innovations are higher, and that they are more likely to expect an increase their investments in innovation.Conversely, there is no similarly clear pattern for firms with other strategic orientations such as a customer-centric or a cost-centric strategy.
The remainder of the paper proceeds as follows: In the next section we provide a brief overview of the literature on economic crises and innovation, and we develop in more detail the research questions this paper addresses, followed by our hypotheses regarding the determinants of crisis-induced innovation.We then continue with a methodological section where we start off by describing the empirical setting, our data and sample, and how the variables are measured.The results of our analyses are reported in Section 4 followed by a discussion of the implications of our findings.We conclude by discussing limitations and suggestions for future research.

Innovation persistence during recession
Given that innovation is widely considered an engine of economic growth and welfare, it is no surprise that the effects of recessions on innovation has generated a lot of interest among scholars, policymakers, and managers (Archibugi 2017, Barlevy 2007;Comin and Gertler 2006;Geroski, Van Reenen, and Walters 1997;Guellec and Wunsch-Vincent 2009;Nuño 2011).There are two obvious questions at the core of this issue; the first is about the aggregate effects, and the second is about the variation underneath these aggregate effects.
The aggregate question is about how aggregate innovation investment and innovation output responds to recessions -and why.The earliest contributors predicted that innovation is mainly motivated by how the push from declining profits motivates firms to venture beyond their normal and known products, routines and practices (e.g.Rosenberg 1976;Schumpeter 1939).This line of thinking would predict countercyclical aggregate innovation.Others have countered that innovation is driven by the lure of higher prices and profits, and the associated increases in knowledge investments and funding for innovation (e.g.Kaldor 1981;Schmookler 1954).The prediction from this line of thinking is exactly the opposite, that aggregate innovation should be procyclical.Antonelli and Scellato (2011) argue that these two opposing views can be reconciled by seeing innovation as a disequilibrium phenomenon, where incentives are higher both when profitability is unusually high and unusually low.
In terms of the empirical evidence on aggregate innovation, this debate seems mostly settled.Innovation is a procyclical phenomenon in the aggregate (Archibugi, Filippetti, andFrenz 2013a, 2013b;Filippetti and Archibugi 2011;Paunov 2012;Roper and Turner 2020).Destruction of demand weakens the incentive to innovate, and more binding financial constraints weakens the ability to innovate (Aghion et al. 2012;Campello, Graham, and Harvey 2010;Knudsen and Lien 2014).These negative impulses outweigh the positive stimulus arising from innovating when excess capacity in other activities (e.g.production and sales) lowers the opportunity costs of innovating, and the inertial effects stemming from the fact that innovation activities are associated with larger adjustment costs than most other activities (Aghion and Saint-Paul 1998;Davis and Haltiwanger 1990;Gali and Hammour 1993;Hall 1992).The net aggregate effect is therefore a loss of innovative input and output.Aggregate innovation is, in other words, procyclical and not persistent.
The variation question turns on the realisation that despite these findings about aggregate innovation, a significant minority of firms are indeed persistent innovators (Amore 2015;Antonioli and Montresor 2021;Archibugi, Filippetti, and Frenz 2013a;Cefis 2003;Spescha and Woerter 2019).These firms maintain or even increase their innovation activities during recessions.This begs the question of who these firms are, and what makes them behave differently from the procyclical innovators?Two findings stand out: The first is that the persistent innovators within a crisis can be characterised as either particularly heavy innovators or fast-growing new entrants before the crisis (Archibugi, Filippetti, and Frenz 2013b;Cefis 2003;Geroski, Van Reenen, and Walters 1997).These firms are the ones that place the heaviest emphasis on innovation and will therefore tend to have the strongest innovation capabilities, routines, and competencies.Maintaining or accelerating their innovation agenda is therefore more attractive than for other firms, for both offensive and defensive reasons.
The offensive argument is that strong innovation capabilities will lead them to expect more valuable innovations per unit of innovation spending.Combined with the reduced opportunity costs of innovating in a recession, they face stronger positive incentives than where innovation is a more peripheral or intermittent activity.In addition, because innovation is associated with cumulative learning processes, heavy innovators can by investing through a recession solidify and improve their strategic position by pulling away from competitors that reduce their investments.The defensive argument is related to the high adjustment costs associated with innovation and R&D (Dierickx and Cool 1989;Hall 2010;Li 2011).High adjustment costs mean that scaling up and down these activities and investments are costly.If you downsize your R&D-division, you cannot expect to scale it back up later and quickly reach productivity levels like those before the downsizing.The main point here being that the more refined and productive the innovation unit is, the higher the adjustment costs will tend to be.Accordingly, firms that have sophisticated, state of the art innovation and R&D activities will have particularly strong incentives to avoid incurring them.

Crisis-induced innovation
What the persistence literature downplays is that during recessions and crises the business landscape changes quite profoundly.While this is generally the case, it has been particularly so during the current COVID-19 recession.Demand evaporates or shifts, supply chains get disturbed, distribution models are left unworkable or even prohibited, employees start working from home, and so on.Firms may deal with such upheavals in different ways.They can cut costs, hunker down, and wait for the storm to pass.Or they can attempt to innovate their products, work processes, target markets, or distribution models to make the best (or avoid the worst) of the new situation.
For some firms, this merely means advancing an existing agenda with respect to innovation and development, but it seems unlikely that this is the general case.Rather, many firms will be forced to consider unexpected innovations in response to unexpected conditions.These are innovation activities that would not have been undertaken in the absence of a recession or a crisis.We label innovations attributable to a recession or a crisis 'crisis-induced innovation'.This can in principle be both accelerating planned innovations and innovating in ways or areas that were not planned.The key element is that it represents a shift in a firm's innovation behaviour relative to pre-crisis intentions and plans.
Thus, innovation is a tool firms can use to soften the impact of a recession, to make the best of bad circumstances, or even to exploit new opportunities created by the change in the competitive context.In other words, a recession or a crisis creates a need for adaptation, and innovation is one way of adapting.It is not the only way, though, and for some firms it may not be the best way either (Knudsen et al. 2022).Innovation is never costless, and the outcome is always uncertain.This is true for innovation under normal circumstances, and there is no reason to think that it would not be true for crisisinduced innovation.For some firms, the expected costs will therefore outweigh the expected benefits, and they will prioritise other tools for adaptation and crisis management.
Given its motive of adapting to an unexpected and changed business landscape, we expect that crisis-induced innovations differ from ordinary innovation.Since it is a crisis response, it is likely to be heavily, but not exclusively, necessity based.Furthermore, we believe that the situation will tend to call for innovations that are preceded by a shorter planning period, implemented faster, and probably also less capital intensive.This in turn suggests that they are more likely to be innovations that are new to the firm rather than innovations that are new to the world, and one might expect that the innovations introduced to be of a more mundane nature 1 -more akin to what Antonelli (2006) calls creative adoption.It also means that firms are likely to favour innovations that do not require heavy investments in the development of new capabilities, but instead favour reallocating, recombining, and reorganising existing capabilities, or to find new partnerships.
Thus, even if crisis-induced innovation is unlikely to be radical and disruptive, it is likely a major part of how the innovation phenomenon plays out in a recession.Importantly, we also believe that crisis-induced innovation is an approach firms will use to varying degrees, and with varying success.Consequently, the general question we ask in this paper is which firms will do more and which firms will do less crisis-induced innovation?

Crisis-induced innovators
We start out by focusing on heterogeneity with respect to how a firm is affected by a recession.It is a well-known fact that firms and entire industries were affected differently by the COVID-19 crisis.More generally, firms and industries differ in their cyclicality.This has been found in every recession ever studied (Petersen and Strongin 1996), and it is likely to affect innovation responses.

Crisis impact and innovation behavior
During a crisis, some firms experience a significant loss of income which pushes them to act.For others, operations continue more or less as usual, while a few firms might even experience windfall gains from a crisis.To shed light on how the impact of the crisis affects innovation behaviour we can draw on a rich literature that addresses organisational performance and innovation (Argote and Greve 2007;Bolton 1993;Cyert and March 1963;Greve 2003;Laursen 2012) and problemistic search (Posen et al. 2018).
Early and seminal work by Cyert and March (1963) postulated that substandard performance fosters innovation.When confronted with poor performance firms are more likely to take risks, search for new opportunities, and adopt new strategies and tactics to turn the tide (Bolton 1993;Greve 2003).Well-performing firms have weaker incentives to take such risks, which means they are more likely to prefer staying on their existing path.During periods of crisis, firms that experience a sudden decrease in 1 Mundane as opposed to innovations that have higher scientific content and represent breakthrough ambitions.
demand (or ability to supply) will experience a drop in income and substandard levels of performance.Since the substandard performance is purely exogenous, some firms might decide to simply sit out the storm while others will want to take immediate action by adapting to the new situation.Firms that are not negatively affected are not subject to the pressure resulting from underperformance.Prospect theory (Barberis 2013;Kahneman and Tversky 1979) makes a related argument by pointing out that decisionmakers have a higher tolerance for risk in the domain of loss than in the domain of gain.In a recession, more firms and decisionmakers are operating in the domain of loss, and according to prospect theory this means that their risk tolerance goes up.Risky search and innovation activities that carry a potential for reducing losses will become more attractive for negatively affected firms, while no such increase will be observed for unaffected firms.Indeed, prospect theory predicts reduced risk tolerance for the firms that are positively affected by the crisis.In combination, this leads us to formulate the following hypotheses: H1: Firms experiencing a decrease in demand (or ability to supply) are more likely to engage in crisis-induced innovation activities.

Innovative capabilities
Even if we controlled perfectly for the severity and nature of the crisis impact, we would probably still observe differences in firms' propensity for crisis-induced innovation.So, which are the firm level characteristics that matter?A reasonable place to start is in the literature on innovation persistence and consider whether findings from this literature are likely to hold for crisis induced innovation also.What this literature has demonstrated is that the firms most heavily engaged in innovation before a recession are more likely to continue with these activities within and after a recession -at least as measured using R&D and innovation survey data (Antonioli and Montresor 2021;Archibugi, Filippetti, and Frenz 2013b;Cefis 2003;Geroski, Van Reenen, and Walters 1997).The reason for such persistence is embedded in two main theoretical arguments: the adjustment cost argument and the competence-based argument.
The adjustment-cost argument (occasionally labelled the sunk cost argument) argues that previous innovation is associated with investments in an innovation infrastructure, including R&D facilities, equipment, specialised human capital resources, and knowledge investments.These investments tend to be sunk and they are usually costly to scale up and down (Antonioli and Montresor 2021;Ganter and Hecker 2013).For example, training R&D personnel often takes a long time, which means that laying off such employees and rehiring when the crisis is over involves a long training period before the new hire reaches full productivity.The more refined and cutting-edge innovation activities are, the more costly they are to switch on and off, and to scale up and down.Under normal times this creates an innovation entry barrier for those who are not innovators and an innovation exit-barrier for those who are proficient innovators.
It also means that the better you are at innovation, the less tempted you are to scale down your innovation activities (Antonioli and Montresor 2021;Ganter and Hecker 2013).
In terms of crisis-induced innovation, as described above, the sunk cost barriers to inexperienced innovators are probably lower because of its mundane and often incremental character.Even so, experienced innovators will presumably have an extra incentive to keep their innovation personnel and infrastructure active and intact, and if necessary, reallocate some attention to more ad hoc innovation to manage a crisis.
The competence-based argument points out that firms with strong previous innovation experience have engaged in a process of knowledge accumulation and have built the necessary capabilities to innovate (Antonioli and Montresor 2021;Ganter and Hecker 2013).Such firms would expect to get more from their innovation efforts both in the short and longer term, and hence they have stronger incentives to innovate.In terms of crisis-induced innovation, the question becomes whether having competencies and capabilities in ordinary innovation (before the recession) means you are more likely to be good at crisis-induced innovation?If crisis-induced innovation is a faster, less capital intensive and less radical version of ordinary innovation, it seems very likely that this is the case.In sum, this means that previous innovators will find innovation more attractive relative to other ways of adapting to a recession -and they will do more of it than firms without previous innovation experience.
If this reasoning holds, we also expect that pre-crisis innovation experience leads to higher expected long-term value from the innovations undertaken in response to the crisis.Certainly, all else equal, any firm would prefer to implement innovation projects that not only respond to the immediate demands of the crisis, but also have high expected value after the crisis.Experience and capabilities in innovation implies that a firm can choose development projects that firms without such capabilities would be unable to realise, and furthermore, that they are likely to get a better outcome from initially similar projects (Hsieh and Tsai 2007).This implies that firms with these characteristics are more likely to pick projects with higher long-term potential, and more likely to get long-term value out of similar projects.We therefore predict that a higher level of pre-crisis innovation experience leads to higher expected post-crisis value from innovations undertaken in response to the crisis.
A third and final expectation consistent with our line of reasoning is that having pre-crisis innovation experience will also affect resource allocation patterns.Specifically, we examine whether firms expect to increase or decrease the resources allocated to innovation.Our prediction is that higher pre-crisis innovation experience increases the likelihood that a firm expects to increase their investments in innovation, and conversely, it decreases the likelihood that a firm expects to reduce their innovation investments.
In sum, this leads to the following three related hypotheses: H2a: More innovation experience before the crisis leads to more crisis-induced innovation during the crisis.
H2b: More innovation experience before the crisis leads to higher expected post-crisis value of a firm's crisis-induced innovations.
H2c: More innovation experience before the crisis leads to higher (lower) likelihood of expecting an increase (decrease) in innovation investments.

Organizational agility
A capability that is likely to increase in importance in the context of crisisinduced innovation is a firm's general ability to adapt and change, which we might label its organisational agility (a term that overlaps significantly with dynamic capabilities) (Janeway 2012;Teece, Peteraf, and Leih 2016;Weber and Tarba 2014).Organizational agility has been defined as: "the capacity of an organisation to efficiently and effectively redeploy/redirect its resources to value creating and value protecting (and capturing) higher-yield activities as internal and external circumstances change" (Teece, Peteraf, and Leih 2016, 17).A crisis, like a recession or a pandemic, is clearly a dramatic -and usually unexpectedchange in the external circumstances.Using innovation as a tool to manage the threats and opportunities brought about by a crisis will also typically involve redeploying and redirecting resources across and within activities.Given that crisis-induced innovation needs to be planned and implemented fast, possibly under severe resource constraints, in a rapidly and unpredictably changing context, firms that have developed a flexible and agile organisation prior to the crisis should possess an important advantage (Rigby, Elk, and Berez 2020).
Having the capacity to 'efficiently and effectively redeploy/redirect' means that the organisational adjustment cost associated with innovating in ways not foreseen before a crisis is lower than for less agile firms.Lower organisational adjustment costs should make crisis-induced innovation more attractive, and we would therefore expect agile firms to do more of it.In addition to ex ante organisational adjustment cost, higher agility entails a better ability to adapt and modify new innovation initiatives underway -as circumstances change or as the firm learns about better ways of realising an idea.An ability to modify, adapt and improve innovation projects underway provides a partial insurance against further unexpected environmental changes, and an ability to get more long-term value out of a given innovation project.In other words, we post the following three (related) hypotheses: H3a: Higher organisational agility before the crisis leads to more crisis-induced innovation during the crisis.
H3b: Higher organisational agility before the crisis leads to higher expected post-crisis value of a firm's crisis-induced innovations.
H3c: Higher organisational agility before the crisis leads to higher (lower) likelihood of expecting an increase (decrease) in innovation investments.

Research context
We use the outbreak of COVID-19 in Norway as the setting for testing our hypotheses.On February 26th 2020, the coronavirus was confirmed to have spread in Norway.As the number of cases rapidly increased, the Norwegian government announced the first national lockdown on the 12th of March.While the health-crisis as measured in deaths and hospitalisation was not as severe as in most other developed countries, the economic consequences of the pandemic was.According to the Norwegian Statistical Office, the COVID-19 pandemic resulted in a 4.7% lower GDP than pre-pandemic estimates (Frederiksen 2021b).This number is approximately equal to the other Scandinavian countries (Frederiksen 2021a).The Norwegian government quickly implemented several countermeasures including easing the regulations on furloughing, reducing employer contributions, providing loan and guarantee schemes, and direct financial support for businesses that experienced a significant drop in revenues.The Norwegian Labor and Welfare Administration (NAV) received over 350.000 applications for unemployment benefits in the 3 weeks after the first lockdown, which amounts to approximately 12.5% of the total labour force.Around 90% of these applications concerned unemployment benefits related to furloughing.

Data and sample
Between 16 November and 13 December 2020, we conducted an online survey distributed among 16.475 Norwegian CEOs.In this study, we limited the sample to CEOs in firms active in the private non-primary sectors, who employed at least five employees.1.697 CEOs participated in the survey, but this number was subject to attrition and missing variables since the number of respondents being used in our final analysis was 872.2 Upon closer inspection there appears to be no bias in terms of industry, overall size categories, liquidity rates or debt ratio.However, our sample has firms that are on average slightly larger and slightly older.In other words, non-responses are somewhat more common among smaller and younger firms.Still, such firms remain a large share of our sample.
In terms of attrition bias, we use Student's t-test and Chi 2 test to examine whether observations in our empirical analysis differ from those observations not included due to missing variables.More specifically, we test the presence of significant differences between the two groups in terms of industry affiliation, firm size, firm age, whether the firm is affected by the crisis (question 1 in the survey), and strategic posture before the crisis (question 2).These tests are reported in the Appendix (see Table A2, Table A3 and Table A4).We identify no attrition bias according to whether firms were affected by crisis or not, or the age of the firm, but some minor biases with respect to industry and size.In terms of industry, we identify slightly lower attrition rates in Retail and Wholesale, and Administrative and Support Services and slightly higher attrition in Manufacturing of Food Products, Beverages and Tobacco products.In terms of size, we see slightly higher attrition rates in the size category 10-24 employees, but t-test on average size does not show any significant differences.Finally, there is only a minor attrition bias in one of the pre-pandemic strategies, the one on higher customer service.Even though we observe some attrition bias, we evaluate this bias as being minor.

Dependent variable
For our dependent variables, we create different measures that provide us with complementary insights on the crisis-induced innovation responses by firms.Firms that find innovation responses more attractive should do more of it during the crisis (i.e. higher innovation output), they should expect more from the innovations they undertake (i.e. higher expected post-crisis value), and they should be more prone to increase the resources allocated to innovation (i.e. higher expected innovation investments).In other words, see these different dependent variables as complementary manifestations of the same underlying phenomenon.
To measure innovation output, we draw on survey data that directly asked the respondents to what extent their firm, in response to the COVID-19 pandemic, had: (i) developed new products and services; (ii) developed new or improved processes; (iii) targeted new customer groups; and/or (iv) developed new logistical solutions.Based on this measure, we create several crisis-induced innovation output measures.First, we create a dummy variable that indicated whether the firm had introduced any crisisinduced innovations in response to the COVID-19 pandemic.This variable receives a value of 1 if the CEO responded 'to some extent' or 'to a large extent' to any the innovation output categories.
While this measure would assist in addressing our hypothesis, we are aware that there might be variation among the different type of innovations.Consequently, we create separate measures for each of the four innovation categories, which we analyse in a separate regression analysis.Here, we follow the same procedure and create four separate dummy variables for innovating 'to some extent' or 'to a large extent' on new products and service, new or improved processes, entering new markets, or significantly changed logistics.
To test hypotheses 2b and 3b, we rely on a survey question where we asked Norwegian CEOs if they agree with the following statement using a 5-point Likert scale: 'The innovations will be important for us even after the COVID-19 pandemic is over.'When they answer positive to this question, we create a dummy variable with the value 1 and 0 otherwise.Finally, to address hypotheses 2c and 3c, we create a measure of expected changes in innovation investments compared to pre-crisis intentions -notably as a consequence of COVID-19.More specifically, we asked respondents in the survey whether they expected a change in innovation investments using a 5-point Likert scale, which varies from 'a large reduction' to 'a large increase'.Based on this survey question we created two innovation investment measures: (i) a dummy variable that indicates whether they expect an increase in innovation investments; and (ii) a dummy variable that indicates whether they expect a decrease in innovation investments.

Independent variables
Our main independent variables are firm-specific characteristics we have theorised would be drivers of innovation responses.One is the extent to which a firm experiences a decrease in demand or a decrease in the ability to supply.To measure this decrease we created two proxies for being negatively affected by the COVID-19 pandemic.First, we created a dummy variable called 'reduced capacity', which gets the value 1 if the firm, due to the pandemic, is not fully utilising their capacity, for example reduced opening hours, reduced service offering or reduced activity level.The dummy variable also receives the value 1 if they have temporarily closed their business due to the pandemic.Second, we create a dummy variable called 'downsized' which gets the value 1 if the organisation has downsized due to COVID-19 pandemic, either via layoffs or furloughing.
A second set of variables concerns the pre-pandemic strategic emphasis of a firm.We postulate that differences in a firm's capabilities and organisational characteristics are reflected in the strategies a firm pursues.Firms competing on prices and cost will tend to have different capabilities and characteristics than firms competing on innovation, or on customer relationships and branding.If true, this should be reflected in a clustering of competitive behaviours consistent with key differences in underlying capabilities and characteristics.To identify such clusters, we asked respondents about their emphasis on 14 different competitive parameters in their competition with their closest rivals before the onset of the recession.Afterwards, we conduct a principal component analysis (PCA) to condense the strategic orientation into manageable categories.Based on the PCA with varimax rotation, we identify three components with an eigenvalue greater than 1.According to the factor loadings, we label these components as innovation-centric, customer-centric and cost-centric strategies (see Table 1), which confirms our assumed clustering of firm strategies and is also consistent with previous research (e.g.Knudsen and Lien 2014).While we have not developed distinctive hypotheses about customer-centric and cost-centric strategies, we nevertheless include these measures in the analysis to contrast with an innovation-centric strategy.
In addition to a firm's pre-pandemic strategy, we also create a proxy for organisational agility.By this we mean the ability of the organisation to quickly respond to changes in their competitive environment.In the survey, we included a question about how the firm performed compared to its competitors with respect to its ability to rapidly react to new opportunities and threats (before the crisis).If the respondent answered that their ability was stronger or much stronger, we classified the firm as an agile organisation.

Control variables
From the publicly available financial data, we are also able to extract series of control variables.First, we include a measure of the size of the organisation.Here we create six size categories: 5-9 employees, 10-24 employees, 25-49 employees, 50-99 employees, 100-249 employees and more than 250 employees.Second, for industry controls we take a point of  departure in the A38 NACE rev 2 industry classification.When confronted with smaller industry classes we merge them with the most related industry class. 3 We create a measure of geographic location, in this case focusing on the level of the county. 4he literature on innovation persistence has found that financing constraints matter.Thus, we also include two financial measure that previous research has found to affect the investment behaviour of firms, including investments in innovation.These measures are the pre-pandemic liquidity and debt ratios of the firm.We obtain this information from publicly available financial data from 2019 which we merged with our survey data.
In Table 2, we provide our descriptive statistics including a correlation table of all the variables in our empirical analysis.As mentioned above, we only present descriptive statistics for the observations that we use in at least one of our analyses, meaning that we include 872 observations in this table.Clearly, the crisis did spur innovation responses, as more than 50% of respondents indicated to have developed new products and services, new processes, entered new markets and/or developed new logistic solutions in response to the COVID-19 pandemic.Out of the firms that introduced a crisis-induced innovation to a large or some extent, 82% indicated that they expect that these innovations have long-term value.In terms of expected changes in investments, 27% expect an increase while 12% expect a decrease.Forty-seven per cent of our respondents answered that they regard themselves as more responsive to opportunities and threats than their competitors and are thus classified as agile.
Instead of only limiting to reduced capacity and downsizing, we combine these variables in four categories.More than 43% of our sample has experienced neither.Just over 28% stated that they have downsized, but operated under full capacity, while nearly a quarter indicated that they both downsized and operated with reduced capacity.A small share indicated that they operated with reduced capacity but did not downsize their labour stock.

Empirical strategy
Given the binary nature of our dependent variables, we estimate several probit models.More specifically, we estimate the effects of our independent variables on the probability that a firm (i) introduces a crisis-induced innovation; (ii) expects that the innovation has a long-term value; and (iii) expects to increase or decrease innovation investments in the future, which in our specification is identified by y i � .The binary probit model equation takes the following form: where xx is a vector of covariates which includes our main variables of interest and our control variables.β is the corresponding coefficient vector, and the error term (εε) captures the unobserved factors influencing the innovation decision.The latent variable y i � is a binary variable for any of our innovation measures, which we specify: Our estimates cannot be interpreted similar to a standard OLS regression.For that reason, we calculate the marginal effects in all our models.This allows us to provide a meaningful interpretation on the size effects within and between the different models.
In addition, we must recognise that when separately estimating new product and service innovations, new processes, entering new markets, and new logistical solutions we run into the issue that these are not mutually exclusive categories.On the contrary, these innovation choices might be interrelated and thus correlated.Indeed, the correlation table (Table 2) demonstrates that a moderate correlation exists between these innovation categories.For that reason, when analysing the determinants of these crisisinduced innovation categories, we estimate a multivariate probit model.Like a seemingly unrelated last square regression, this specification allows for the error term to be correlated across different models and thus account for the correlation between the different dependent variables.

Crisis-induced innovation
The results of our probit estimation of the probability that a firm introduced a crisisinduced innovation are presented in Model 1 in Table 3.We observe that both our reduced capacity and downsize variables are associated with a higher likelihood of innovating.More specifically, firms that experience only reduced capacity utilisation have an increased predicted probability of introducing a crisis-induced innovation of 7,5% (0.075).Firms that only downsized, have an increased predicted probability of 16.8%, firms that downsized and experienced reduced capacity utilisation have an increased predicted probability of introducing a crisis induced innovation of 22.4%.Consequently, our empirical evidence supports our first hypothesis.
When we move to our innovation experience and agile organisation variables, hypothesis 2a and 3a, respectively, the model demonstrates that having a pre-pandemic innovation-centric strategy is associated with an increased probability of crisis-induced innovation.Interpreting marginal effects of principal component loadings is not straightforward.In Table 2, we provide the marginal effects, but it is easier to grasp the meaning of a one standard deviation (i.e.1.89, see Table 2) higher loadings on innovation-centric focus.This is associated with 9.4% increase in the predicted probability of innovating.This confirms Hypothesis 2a.We also observe that organisational agility is associated with a 12.9% increased predicted probability of crisis-induced innovation, thereby providing support for Hypothesis 3a.
We would like to add some resolution to this picture.In constructing our dependent variable of crisis-induced innovation, we relied on multiple measures of innovation output (products and services, processes, enter new markets, and new logistics).There is no reason to expect differences in the direction of effects across these innovation categories, but one might speculate that the size of the effects vary.Consequently, we run a separate analysis of these underlying innovation output categories.However, firms  might innovate in all these categories simultaneously, so we can expect a relative strong correlation between them.
As explained in our estimation strategy, this is the reason for running a multivariate probit model.The estimates are presented in Models 2-5 in Table 3.The relation between reduced capacity and crisis-induced innovation is strongest for product innovation and entering new markets, with increased predicted probabilities of respectively 16.4% (Model 2) and 13.9% (Model 4).New work processes seem to be more sensitive to downsizing, with an increased probability of 20.9% in the case of only downsizing (Model 3).This might be explained by a need to modify work processes when working with a reduced labour stock.
Looking at our strategy and organisation variables across different innovation categories have we first of all observe that an innovation-centric strategy is positively associated with the likelihood of innovating in all categories, particularly process innovations.A one standard deviation increase in the factor loading increases the likelihood of process innovations with 12.3%.New products and services comes in second, with an increase in probability of 9.9% (again, for a one standard deviation increase).Also for organisational agility, we find that this is associated higher predicted probability of innovating across all innovation categories, in particular new products and services (increased predicted probability of 10.7%).Thus, we also confirm Hypothesis 2a and Hypothesis 3a on detailed specifications of our crisis-induced innovation measures.

Expectations
The remaining hypotheses all address expectations, both the long-term (post-crisis) value of the crisis-induced innovations and expected changes in innovation investments (Table 4).Firms with an innovation-centric strategy (increased predicted probability of 6.4% following a 1 SD increase) and those that are more agile (increased predicted probability of 10.2%) expect their innovations to be valuable when the crisis has passed (see Model 6).Consequently, not only are firms with pre-existing innovation capabilities and organisational agility better able to innovate, their CEOs also expect to extract more long-term value from the innovations they launch.Consequently, we also find support for Hypothesis 2b and Hypothesis 3b.
When testing Hypothesis 2c and Hypothesis 3c, we observe that firms with an innovation-centric strategy are more likely to expect an increase in innovation investments.From the marginal effect in Model 7 the predicted probability increases by 12,9% for a one standard deviation increase in the PCA score.Simultaneously, firms with an innovation-centric strategy are also less likely to expect a decrease in innovation investments (Model 8), as the predicted probability decreases by 3%.More agile organisation are also associated with an increased probability of expecting higher levels of innovation investments (8.1%).This means that we also find support for hypotheses 2c and 3c.

Robustness test
To test the robustness of our main findings, we run several models on sub-samples on specific industries (cfr.Table 5) and size categories (cfr.Table 6) and replace the  empirical derived strategy measures from the PCA with simplified survey-based measure for innovation-, customer-, and cost-focus (Table 7).When running our industry subsamples, we divide the sample into four industry groups, the choice for these categories is motivated by the size of these industries in the sample.We do observe some sectoral differences.In Model 9, we focused on Retail and Wholesale being the largest industry category.The patterns we observe in this model are much in line with the findings or our main analysis, on all our independent regressors.In Model 10, which focuses construction, we observe that an innovation-centric strategy positively contributes to the predicted probability of introducing crisis-induced innovations.This effect is larger than the one found in our main analysis.The effect of being an agile organisation, however, disappears.In knowledge-intensive industries, we observe the reverse (see Model 11).The effect of previous innovation experience appears to be strongest in manufacturing industries (Model 12), while having a customer and cost focus seems to be negatively associated with introducing crisis-induced innovations.In this category, an agile organisation does not seem to result in higher predicted probability of innovation.
When it comes to different subsamples in size, the results or our main analysis are largely confirmed.However, we see that previous innovation experiences have stronger effects for larger firms.An agile organisation increases the predicted probability of introducing crisis-induced innovations in all size subsamples, but more so in microsized (Model 13) and medium-size enterprises (Model 15).
In Table 7, we replace our PCA measure of innovation-centric, customer-centric and cost-centric strategies with positive responses to the survey items on innovation and R&D, high customer service, and low price.The choice of these items was motivated by their close resemblance to our PCA measures.We observe that that our previous innovation experience measure has a significant positive effect on the predicted probability to introduce a crisis (Model 16), to expect post-crisis value form the innovation (Model 17) and expectations of an increase in future innovation investments (Model 18).Contrary to our main model, we observe that higher customer service is also associated with higher predicted probability to introduce crisis-induced innovations, but no on any of the other measures.

Discussion and conclusion
We have argued that there is more to innovation in recessions than whether firms cut, maintain, or accelerate their pre-recession innovation agendas.This is what the persistence literature mainly addresses.Our focus has been to open a discussion about how firms use innovation as a tool to manage the unexpected and adverse changes that occur in such periods.We have offered some glimpses into the use of the innovation tool, and what creates variation in its deployment.
One finding is that a crisis period -such as the one following the COVID-19 pandemic -seems to drive many firms to innovate.Most firms in our sample have engaged in crisis-induced innovation, at least to some extent.Apparently, then, the phenomenon of crisis-induced innovation is quite widespread, and not limited to the firms that were experienced innovators before the onset of the crisis.Since crisis-induced innovation is mostly undertaken to handle negative fallout from the recession, and since firms need and want the associated benefits to appear quickly, crisis-induced innovation will typically be innovation projects that can be realised fast.This furthermore implies that they are systematically less likely to involve heavy investments in physical or human capital.In other words, firms are unlikely to try to manage a recession by developing radical and disruptive new technologies, or venture into something that requires a long program of building entirely new sets of capabilities.We are talking about reaching for low-hanging fruit that might help, and probably mostly about doing things that are new to the firm rather than new to the world.
We have also found systematic differences in how firms used the innovation tool during the recession.Firms that were adversely affected are generally more motivated to innovate.However, it also appears that there is a somewhat complicated relationship between innovation and downsizing.Overall, it might seem likely that downsizing makes a firm less likely to innovate, and that a firm that innovates is less likely to downsize.However, firms that do downsize appear to engage in innovation activities, but this seems mainly to involve developing new work processes, possibly to create a work-around of the fact that they have downsized.When it comes to other types of innovation, removing excess capacity appears to be a substitute for using it as a motivation for innovation.
It is also quite clear from our data that innovation experience before a recession has a strong impact on how much it is used as a tool during a recession.A strategic emphasis on innovation leads to higher investments in innovation activities, more innovation output of all four types, and a more optimistic view of the long-term value of the crisisinduced innovations undertaken.It seems that those with innovation experience are better able to find crisis-induced innovation projects that have long-term value and simply that more projects are considered attractive for a firm with pre-existing innovation experience.So, while a crisis leads more firms to innovate, there is a sense in which the key finding from the innovation persistence literature is upheld for crisis induced innovation: Innovation expertise matters.
Another finding is that organisational agility also matters.This is not surprising given the premise that we are talking about innovations that need to be implemented fast, probably to a large extent by reconfiguring existing resources and capabilities.Being a flexible organisation with experience in quickly adapting to new threats and opportunities should (as we indeed find) provide important advantages in setting up and implementing the required changes.

Limitations
Our study revolved around innovation-responses of Norwegian firms during the COVID-19 crisis.This raises the question of whether our findings are specific to either Norway or to the COVID-19 recession.We believe that the risk of the latter is higher than the former.The impact of COVID-19 on the Norwegian economy was -as we have shown -quite similar to its impact in other developed economies.It seems more reasonable to question to what extent our findings are generalisable to other crises.The causes and the nature of the COVID-19 crisis were arguably different than, say, the financial crisis.Still, it seems that any crisis that has adverse effects on many firms, and where innovation might be a tool to manage its impact, should result in what we have called crisis-induced innovation.Nevertheless, we strongly recommend replications and extensions to future crises, and if data allow also on previous crises.Replication in other geographical contexts would also be useful.
Another set of limitations are the inherent constraints associated with the survey method.For one, there are concerns over response biases.As discussed in the method section, we believe that while our sample is not completely free from response-and attrition biases, these seem too small to seriously distort our findings.Our survey data is also cross-sectional.This does not provide us with clear cut time variance and limits our ability to clearly determine causality.We have attempted to address this problem by specifically asking for information referring to the pre-COVID situation when addressing the question of capabilities, and by asking specifically about responses and decisions that were taken in response to the crisis.This should mean that reverse causality for these hypotheses is unlikely to be a problem.Another methodological challenge occurs because the impact of the crisis is seen as an incentive for firms to innovate, but it is also capable of reducing the severity of the crisis impact.Attempts to address this issue, for example through the use of an instrumental variable approach was unsuccessful.The risk that crisis impact is severely affected by the innovations undertaken in response to the crisis is also somewhat limited by the fact that our survey was conducted quite early (while the COVID-19 crisis was in its second wave, less than a year after the onset of the crisis), and the positive effects of innovations will often take some time to materialise.Still, we acknowledge that this remains an imperfectly resolved issue.Furthermore, we cannot rule out that omitted variables have influenced our findings.Retrospective survey data may also suffer from recall bias.We minimised this problem by conducting our survey while the COVID-19 crisis was still fresh in the minds of our respondents.While the relatively short time span covered is helpful in terms of ensuring accurate recall about responses and strategies before the crisis, it also means that we have not been able to observe the real long-term impact of the innovations.Consequently, it is not possible to assess whether innovators are more resilient or otherwise successful than their non-innovating counterparts.This is obviously a relevant question to address in future research.
A final limitation is the measurement of innovation.We do not have detailed information on how incremental or radical the innovations are.Given the context driving crisis-induced innovation, we have argued that these innovations are likely to be more incremental, less time consuming and less capital intensive.We admittedly base this on theoretical reasoning, not observations.It would clearly be useful to establish if our reasoning holds empirically, and it would also be useful to learn more about what makes firms launch more (or less ambitious) crisis-induced innovations.

Practical implications
When managers decide which strategy to pursue, they also shape which capabilities their firm will develop.Which capabilities they develop will in turn heavily influence which tools they turn to in a recession or other kinds of unanticipated shocks.Those that emphasise innovation and agile organising will be the ones that use the innovation tool the most, and those that emphasise costs, price, and efficiency will be the ones that emphasise it the least.The potential value of using the innovation tool to solve unexpected problems, crises, and shocks, should therefore be factored in already at the strategy formulation stage.It further suggests that if and when firms believe their environment will become more prone to unexpected shocks, cyclical turbulence, or other types of hard to anticipate crises, they should consider increasing their emphasis on developing these two characteristics.Our findings are also potentially useful for managers as a basis to foresee how their competitors, suppliers, business customers, and complementors are likely to react once an environmental shock hits.
For policymakers, it is presumably important to understand how different support schemes affect the behaviour of the firms they target.Our findings reveal new information about how innovation responses are likely to be influenced by different support schemes.We find that financing is not as big a constraint for crisis-induced innovation as it has been found to be for ordinary innovation.The constraints seem to be more rooted in lack of innovation experience and the necessary organisational perquisites.We also find evidence that downsizing leads to lower expected investments in innovation.Support schemes that keep human capital slack in the firm, and thereby maintain a firm's ability to reallocate capacity to innovation and other types of development work will stimulate crisis-induced innovation.Support schemes that merely compensate employees for not going to work will result in less innovation and possibly weakened or forgone innovation capability in the wake of a crisis (Klein et al. 2020).A relevant issue for policymakers is therefor to consider how retention schemes can be implemented in a way that does not lead to human capital that could have helped innovate becoming passive.

Research implications
The big question that towers over all these findings is whether crisis-induced innovation is worthy of scholarly attention at all, or whether it is a trivial and fleeting phenomenon that can be ignored in favour of more important matters.It is perhaps not entirely surprising that we believe that this is an important aspect of the wider innovation phenomenon that warrants attention, and we also believe it raises many questions that we have been unable to answer here.We believe the phenomenon is important because a major part of the innovation behaviour that takes place in a recession is presumably overlooked or severely underemphasised in existing empirical work.What we call crisisinduced innovations are for the most part unlikely to have a strong scientific and technological basis.Consequently, they are not likely recorded in patent statistics.Much of it is probably not recorded as R&D-expenses, either (Abreu et al. 2010;Nesta 2007).We even suspect that much of this is not fully captured by data from periodic surveys such as the Community Innovation Survey.Some of these innovations might have been forgotten by the time such surveys are distributed, because their purpose has been served by helping the firm manage through a difficult period and they are subsequently abandoned.They may also no longer be considered innovations because the innovations are to a large extent tied to improvements of ongoing practices (Saidi, Thune, and Bugge 2021).Finally, the majority of firms in our study are active in services and non-high-tech industries, a sector where one traditionally observes high levels of hidden or invisible innovation (Djellal and Gallouj 2010;Gallouj et al. 2011;Nesta 2007).Engaging in efforts to understand this innovation phenomenon also aligns with the call to include innovation activities that move beyond tech-savy, R&D intensive to include less glamorous innovations that for many firms are more appealing and realistic (Martin 2016).
The ability to use innovation -even if it is a mundane form of innovation -as a tool to manage the effects of a crisis can have positive effects if it helps mitigate the negative effects it brings, or if it makes the selection effects that takes place in recessions more efficient.It can also have positive effects by giving non-innovators the impetus to gain some early innovation experience, which might lead them towards more ambitious innovation activities later.It can of course also have negative effects.Crisis-induced innovation might mainly crowd out innovation with larger potential in the long run, or it might simply be a costly distraction for most of those who engage in it.At this point in time, we still do not have data on whether (or for who) these innovations bring material long run benefits in terms of survival, growth, or profitability -or not.But we do think our data shows that there are significant and interesting changes in innovation patterns taking place during a recession such as the COVID-19 crisis, and that these changes have mostly flown beneath the radar of the existing innovation literature.

Q2.4.3 To what extent do you agree with the following statements
Completely disagree

Somwhat disagree
Neither agree nor disagree

Somewhat agree Completely agree
The innovations will be important for us even after the COVID-19 pandemic is over.bold values significantly different on 5 percent confidence.
.05 highlighted in bold.

Table 1 .
Principal component analysis on competitive strategies (varimax rotation) and correlation.

Table 2 .
Descriptive statistics and correlation table.

Table 3 .
Probit models on crisis-induced innovation.

Table 4 .
Probit models expected post-crisis value of innovations, and increase and decrease in innovation investments.

Table 7 .
Robustness probit regression using survey-based measure of pre-pandemic strategies.

Table A3 .
Chi-square test on industry, size, and covid impact.