Introduction
A firm’s ability to innovate is one of the factors that characterize its sustainability and potential for growth (Dukeov, Pekka Bergman, Heilmann, & Nasledov, Reference Dukeov, Pekka Bergman, Heilmann and Nasledov2020). The adoption of innovation entails organizational adaptation for the purpose of facilitating the fulfillment of goals, especially under the intense competition conditions, rapidly changing markets, scarce resources, and increasing demands for higher-quality products and services (Boyne, Farrell, Law, Powell, & Walker, Reference Boyne, Farrell, Law, Powell and Walker2003; David Osborne, Reference David Osborne1993; Pollitt & Bouckaert, Reference Pollitt and Bouckaert2011). Despite numerous studies on the relationship between innovation and organizational performance, different studies have reported different and conflicting reports for this relationship (Hu, Hu, & Parsa, Reference Hu, Hu and Parsa2015; Prima Lita, Faisal & Meuthia, Reference Prima Lita, Faisal and Meuthia2020). Although the overall view of this relationship is important, due to the growing globalization of industries and businesses and the development of virtual teams and international companies, understanding the difference in intensity of this relationship in different societies is very much needed. Moreover, creating a comparative perspective for this relationship with regard to the types of innovation is necessary for future research and applications. Furthermore, considering the rapid developments of societies, it seems necessary to examine the changes in this relationship in recent years.
The studies on the impact of innovation on organizational performance have used different methods, tools, and samples. A study conducted in China in 2014 confirmed the relationship between innovation and organizational performance in its universities with a high effect size (Cheng, Reference Cheng2014). Matjaž Maletič et al. demonstrated the positive and strong relationship between innovation and organizational performance (Maletič, Maletič, Dahlgaard, Mi Dahlgaard-Park, & Gomišček, Reference Maletič, Maletič, Dahlgaard, Mi Dahlgaard-Park and Gomišček2016). Another study was conductedin Spain and estimated a medium effect size for the relationship between innovation and organizational performance in technology companies (Martín-Rojas, Garrido-Moreno, & García-Morales, Reference Martín-Rojas, Garrido-Moreno and García-Morales2020).
Han and Huang (Reference Han and Huang2012) analyzed the effect of innovation on the organizational performance and concluded that increased innovation has a slightly significant effect on their organizational performance (Han & Huang, Reference Han and Huang2012). Although previous studies reported the positive effect of job satisfaction on job performance, Hsin Hui et al. concluded that marketing innovation has a negative relationship with organizational performance (Hu et al., Reference Hu, Hu and Parsa2015). Moreover, in some other studies, an insignificant relationship between innovation and organizational performance was reported (Ilmudeen, Bao, Mubarak Alharbi, & Zubair, Reference Ilmudeen, Bao, Mubarak Alharbi and Zubair2020; Prima Lita et al., Reference Prima Lita, Faisal and Meuthia2020). Considering the difference in the results of previous researches, it has been important for researchers to conduct a meta-analysis to achieve a unified result. Liao, Liu, and Liu (Reference Liao, Liu and Liu2021) examined the relationship between environmental innovation and performance from January 1985 to December 2019 using English databases (Liao et al., Reference Liao, Liu and Liu2021).
Feng, Ma, and Jiang (Reference Feng, Ma and Jiang2020) conducted a meta-analysis between service innovation and performance from 1986 to 2019 in English databases and their results illustrated that service innovation has a negative relationship with organizational performance (Feng et al., Reference Feng, Ma and Jiang2020). Thao Nguyen, Huang, and Tian (Reference Nguyen, Huang and Tian2021) examined the relationship between open innovation and performance based on a comprehensive data set of 2,377,123 firms and sub-firm units in 171 studies published from 2003 to 2018 and showed that open innovation activities, including user participation, significantly contribute to firm performance (Nguyen et al., Reference Nguyen, Huang and Tian2021). Damapanpour, Sabat, and Evan (Reference Damapanpour, Sabat and Evan1989) examined the relationship between type of innovation and performance. They focused on the administrative and technical innovations. Damanpour used public libraries data in six northeast states in the United States data during the 1970s (Damapanpour et al., Reference Damapanpour, Sabat and Evan1989). Bowen, Rostami, and Steel (Reference Bowen, Rostami and Steel2010) investigated the relationship between innovation and firm performance meta-analytically using 158 effect sizes from 55 empirical studies in both temporal sequences, and the result illustrated that the effect size of this relationship is 0.16 (Bowen et al., Reference Bowen, Rostami and Steel2010).
The affluence of research conducted on this relationship demonstrates its importance (Bowen et al., Reference Bowen, Rostami and Steel2010; Feng et al., Reference Feng, Ma and Jiang2020). The difference in the results of past researches has determined the necessity of conducting meta-analytical researches in this field. However, previous meta-analyses have been focused on a specific type of innovation or performance (Liao et al., Reference Liao, Liu and Liu2021) or were published years ago (Damapanpour et al., Reference Damapanpour, Sabat and Evan1989) that do not provide a comprehensive and comparative view, which given the changes in societies in recent years, it is necessary to re-examine this relationship meta-analytically. Furthermore, this study presents a novel approach compared to other meta-analyses in the field. Unlike previous studies that have categorized different types of innovation, this research has avoided the categorization of innovation types to ensure the independent examination of each type of innovation. By adopting this approach, the analysis aimed to maintain the separation and distinctiveness of the various types of innovation, avoiding potential confounding effects that could arise from their integration. In addition to creating a comprehensive view of this relationship, this study has provided researchers with an understanding of the difference in this relationship in different fields by identifying variables such as country, continent, types of innovation, and the year of publication as moderator.
The results of this study can enable the managers of different industries to have a more comprehensive view of the subject and have better planning to improve their organization’s performance through innovation. Furthermore, the results of this research are useful for researchers especially in Spain, China, Iran, Brazil, South Korea, and the UAE to investigate the reason for the difference in this relationship in those countries and to localize the results of international research. Moreover, clarifying the trend of changes and fluctuations in this relationship during recent years is one of the other achievements of this research, which can be useful for managers and researchers. This study aims to integrate and analyze the results of published studies on the relationship between innovation and organizational performance from January 2012 to January 2022 using meta-analysis and achieve a consistent result with high statistical power. In ‘Literature review’, the review of the literature is presented. The methodology of the meta-analysis is presented in ‘Methodology’. Results and reports of searching the databases, software outputs are presented in ‘Results’. Discussions and limitations are provided in ‘Discussion and implications’. Finally, the conclusion is presented in ‘Conclusion’.
Literature review
Innovation
Innovation, defined as the creation or significant improvement of a new product (product or service), process, new marketing method, or organizational method, is seen as a tool to increase corporate performance and competitive advantage (Shanker, Bhanugopan, van der Heijden, & Farrell, Reference Shanker, Bhanugopan, van der Heijden and Farrell2017). Research demonstrated that employees’ innovation positively contributes to the success of an organization (Douglas & Sutton, Reference Douglas and Sutton2010). In another study on small- and medium-sized enterprises in the manufacturing sector of Malaysia, a significant relationship between innovation and organizational performance was found (Zakaria, Chew Abdullah, & Zien Yusoff, Reference Zakaria, Chew Abdullah and Zien Yusoff2016). In a separate study conducted on technology managers in Spain, the relationship between innovation and organizational performance with high impact size was confirmed (Martín-Rojas et al., Reference Martín-Rojas, Garrido-Moreno and García-Morales2020).
Organizational performance
Performance is a multidimensional concept that defines the position of the organization relative to its competitors (López-Nicolás & Meroño-Cerdán, Reference López-Nicolás and Meroño-Cerdán2011). Organizational performance refers to the process of explaining the effectiveness of prior actions (Neely, Gray, Kennerley, & Marr, Reference Neely, Gray, Kennerley and Marr2006). Therefore, organizational performance is related to how much an organization achieved its goals (Li, Ragu-Nathan, Ragu-Nathan, & Rao, Reference Li, Ragu-Nathan, Ragu-Nathan and Rao2006). Currently, evaluating the performance of organizations, which involves the process of quantifying the effectiveness of previous activities, gains significance (Tangen, Reference Tangen2004). By evaluating performance, organizations can understand the current situation and examine future challenges (Wang & Noe, Reference Wang and Noe2010). One of the key factors determining organizational performance is divided into two categories: the factors related to internal and external environment. Internal factors related to internal environment encompass strategy, structure, leadership, development and innovation, information technology, performance measurement, quality, and staff. External factors related to external environments refer to customers, competitors, and suppliers (Gavrea, Liviu, & Stegerean, Reference Gavrea, Liviu and Stegerean2011; Mokhber, Khairuzzaman, & Vakilbashi, Reference Mokhber, Khairuzzaman and Vakilbashi2018).
Innovation and organizational performance
Today, innovation is considered as an important mechanism for achieving competitive advantage and survival in global business (Ayoko, Reference Ayoko2021; Sadikoglu & Zehir, Reference Sadikoglu and Zehir2010). The ability to innovate provides organizations with a strategic orientation to achieve a sustainable competitive advantage (Wiklund & Shepherd, Reference Wiklund and Shepherd2011). Different organizations and even countries seek to rely on innovation to increase productivity and improve the economic conditions which contribute to the growing significance of innovation and increasing competition among communities and organizations (Jin, Hewitt-Dundas, & Thompson, Reference Jin, Hewitt-Dundas and Thompson2004). In the relationship between technology-based innovations and competitive advantage is often shaped by organizational competencies that enable firms to exploit the results of their technological assets (Garcia-Morales & Esmeralda, Reference Garcia-Morales and Esmeralda2018). This leads to the first hypothesis.
Hypothesis 1: Innovation relates positively with organizational performance.
Refined articles from databases have been made in different countries and years. Furthermore, cultural background of societies can boost innovation and thus accelerate competitiveness (Petrakis, Reference Petrakis, I.N. and C.2016).
Metcalfe (Reference Metcalfe1998) stated that when the flow of newness and innovations desiccates, firms’ economic structure settles down in an inactive state with little growth. Therefore, innovation plays a significant role in creating the differences of performance and competition among firms, regions, and even countries (Metcalfe, Reference Metcalfe1998). For instance, the study by Fagerberg, Mowery and Nelson (Reference Fagerberg, Mowery and Nelson2004) revealed that innovative countries had higher productivity and income than the less-innovative ones (Fagerberg et al., Reference Fagerberg, Mowery and Nelson2004). The Organisation for Economic Co-operation and Development (OECD)reports pointed out that companies that developed innovations in a more decisive way and rapidly, had also more qualified workers, paid higher salaries, and provided more conclusive future plans for their employees. In fact, the effects of innovations on firm performance differ in a wide spectrum from sales, market share, and profitability to productivity and efficiency (OECD Oslo Manual, 2005). Industry and business administrations in different countries may use different innovations according to the culture and language of the employees. On the other hand, the performance of employees in different countries is not the same. Different results of studies in different countries and years lead to the second, third, and fourth hypothesis.
Hypothesis 2: Continent is a moderation variable for the relationship between Innovation and organizational performance.
Hypothesis 3: Country is a moderation variable for the relationship between Innovation and organizational performance.
Hypothesis 4: Publication year is a moderation variable for the relationship between Innovation and organizational performance.
Innovation encompasses various types, each with different impacts on organizational performance. Cheng Zhu, Isaac, and Edmund (Reference Zhu, Isaac and Edmund2022) examined the relationship between innovation and firm performance, highlighting the importance of innovation for competitiveness and economic growth (Zhu et al., Reference Zhu, Isaac and Edmund2022). The results show that innovation has positive influence on the performance of Chinese enterprises (β = 0.059, p < .01), and suggesting that innovative efforts by Chinese firms have impact on their performance. This is consistent with previous findings by Syed et al. (Rehman Khan et al., Reference Rehman Khan, Yu, Sarwat, Iqbal Godil, Amin and Shujaat2022) and Guo et al. (Guo, Chao, Zhongfeng, & Donghang, Reference Guo, Chao, Zhongfeng and Donghang2020), where it was established that innovation fosters enterprise productivity. Zhon Liao et al. conducted a meta-analysis between environmental innovation and firm performance and reported that the environmental innovation has a significantly positive effect on both firms financial and environmental performance (Liao et al., Reference Liao, Liu and Liu2021). This conclusion is consistent with the results that Ong et al. (Ong, Lee, Heng The, & Bakhsh Magsi, Reference Ong, Lee, Heng Teh and Bakhsh Magsi2019) and Xue, Boadu, and Xie (Reference Xue, Boadu and Xie2019) reported. Changli Feng et al. conducted a meta-analysis and demonstrated that relationship between service innovation and firm performance is significantly affected by the economic region. The results of the meta-analysis have indicated that service innovation has a more significant effect on firm performance in developing regions than in developed regions (Feng et al., Reference Feng, Ma and Jiang2020). The relationship between innovation and organizational performance can be influenced by various types of innovation, leading to the formulation of the fifth hypothesis.
Hypothesis 5: Type of innovation is a moderator variable for the relationship between innovation and organizational performance.
Methodology
Meta-analysis is a statistical method that integrates the results of several independent studies that are ‘combinable’ (Egger, Smith, & Phillips, Reference Egger, Smith and Phillips1997). Qualitative review studies could provide insights into expanding the literature in this area. Qualitative review studies contribute to the development of theories, but do not provide generalizable result. Quantitative and systematic review studies are gaining significance due to providing generalizable results and following a specific goal to prevent bias and introduce various areas for future research studies. Moreover, quantitative review studies could more effectively bridge research gaps (Rowley & Paul, Reference Rowley and Paul2021; Singh, Dhir, Gupta, Mukunda Das, & Sharma, Reference Singh, Dhir, Gupta, Mukunda Das and Sharma2020).
The current study aims to investigate the relationship between innovation and organizational performance. The quantitative procedures in meta-analysis studies eliminate the challenges caused by multiple responses given to a particular question in various studies. Meta-analysis combines the numerical results of several studies, makes an accurate estimation of descriptive statistics to explain incongruities, and discovers moderators and mediators. It enables the researchers to obtain the results that are more accurate and valid compared to what is achieved in a preliminary study or a non-quantitative validity investigation (Rezaian, Reference Rezaian2005).
In the process of conducting a meta-analysis, first thorough and required searches are performed among past studies; next, the information and data are extracted from studies that have been identified based on a specific protocol. Then the results of this extracted information are combined. After analyzing these data, various factors such as homogeneity or heterogeneity of them, biases, and identifying the moderator variable are examined (Borenstein, Hedges, Higgins, & Rothstein, Reference Borenstein, Hedges, Higgins and Rothstein2011). The next step is conducting necessary examinations and analyses, the results of which are reported as effect sizes. The value that represents the magnitude of the relationship between two variables is the effect size.
Moreover, for readers to make the right decisions in scientific terms there must be an appropriate criterion. This criterion is the observed magnitude of the effects, which a quantitative amount that must be reported by scientific processes. The obtained index summarizes effect sizes of findings in meta-analysis (Grissom & Kim, Reference Grissom and Kim2005). Cohen (Reference Cohen1992) defined a criterion for describing the magnitude of the effect size. Basically, for this criterion, the effect size of 0.5 means a large relationship, 0.3 means a medium relationship, and 0.1 means a small relationship. The relationship is also significant and reaches a 95% confidence level when the p-value is less than .05 (Cohen, Reference Cohen1992).
The strategy of searching the literature
A systematic and comprehensive search was conducted on ASCE, ScienceDirect, and Scopus databases. In addition, the Magiran Persian database was searched to prevent linguistic bias. The current study investigated the studies published between January 2012 and January 2022. The keywords searched in this process included ‘innovation’ and ‘organizational performance’. The search field of keywords was title, keywords, and abstract.
Inclusion and exclusion criteria
The current study can be classified as a documentary research study that applied a quantitative approach and a descriptive strategy to comprehensively investigating the studies with a certain hypothesis and obtaining a specific conclusion. The studies were downloaded from four databases. According to the Fig. 1, articles that were overlapping in databases were removed, then qualitative research studies were excluded. In the next step, the studies need to observe the following criteria for data analysis purposes: making a hypothesis or raising a question, using a reliable and valid data collection instrument, having a completely specified population, and using convenient statistical tests. Based on the flowchart illustrated in Fig. 1, 424 studies were identified in the databases. However, 254 studies were missing, and 175 articles were repetitive (a study was repeated three times and other studies were repeated twice), which led to the elimination of 88 studies. In the next phase, 308 qualitative articles were discarded. Then, 631 studies were eliminated since they neither investigated their intended hypothesis nor conformed to the protocol defined in the current study. Finally, 143 studies were selected and the required data were extracted.
Flow diagram for the identification and exclusion of studies.
Data extraction
Based on the aforementioned points, the articles were selected according to the Fig. 1. After the articles were selected, the patterns of the correlation coefficient and the sample size were used for software input insertion. Some articles reported the β value instead of the correlation coefficient. Formula 1 was used to convert it into the correlation coefficient (Peterson & Brown, Reference Peterson and Brown2005).
\begin{equation}{\text{r = 0}}{{.98\beta + 0}}{{.05\lambda }}\;\;\;\left\{ {\begin{array}{*{20}{c}}
{{{if \beta \gt 0\ \lambda = 1}}} \\
{{{if \beta \lt 0\ \lambda = 0}}}
\end{array}} \right.\end{equation}Data analysis process
The data were analyzed by Comprehensive Meta-Analysis Software (CMA2). To this end, the magnitude of the relationship between the two variables is measured by effect size. In addition, Cohen’s coefficient (Reference Cohen1992) was applied to interpret effect sizes (Cohen, Reference Cohen1992). Thus, r < 0.3 indicated a low effect size, 0.3 < r < 0.5 indicated a medium effect size, and r > 0.5 indicated a high effect size. In addition, the current study used the funnel chart and Egger’s regression intercept test (Egger et al., Reference Egger, Smith and Phillips1997) to assess publication bias. Furthermore, Duval and Tweedie’s trim and fill method and the classic fail-safe N were applied to determine the number of missing studies, whereas the test of heterogeneity was used to detect moderator variables. The important results of each test will be presented in the Results section. In addition, the Q-test, I-square, and τ-square procedures were applied to analyze the homogeneity of the studies (Borenstein, Hedges, & Rothstein, Reference Borenstein, Hedges and Rothstein2009; Reference Borenstein, Hedges, Higgins and Rothstein2011).
Results
An introduction of the reason and necessity of conducting this meta-analysis on the relationship between innovation and organizational performance was discussed in the previous sections. A summary of the results of the previous studies and a literature review of the issue were provided. The tests applied in this study were also described, along with the research methodology. The descriptive and inferential statistics are explained in this section. Inferential statistics are also used to describe more accurate information about the articles and to obtain the separated results of analyzes and tests.
The features of the selected articles
As described in the ‘Research Methodology’ section, a specific protocol was used to select the articles. Finally, out of 1,170 articles, 143 articles with 180 data were selected. Selected articles include various statistical populations such as bank employees, telecommunication center employees, automotive industry managers, restaurant employees, hotel staff, university faculty members, public sector employees, SMEs company managers, hospital and medical center employees, IT managers of companies as well as the owners, and employees of manufacturing companies.
The articles were published from 2012 to 2022 and conducted in countries such as China, Iran, UAE, Saudi Arabia, Pakistan, India, Taiwan, Indonesia, Malaysia, South Korea, Germany, Spain, Serbia, Netherlands, Turkey, French, Romania, United States, Brazil, South Africa, Ghana, etc. Questionnaires and surveys were used as data collection tools in these articles. Some selected studies reported more than one data source and sample size due to the type and number of the reported data.
The statistics of the selected articles
As stated in advance, given the systematic review processes, reporting the features of articles is required after identifying them by following the predefined protocol. Then, the Pearson’s correlation coefficient, homogeneity tests, correction fitting by Dual and Tweedie’s method, the Egger’s test and funnel plot, the Rosenthal’s fail-safe N test, and the moderator variable analysis, are described in this section. Moreover, the obtained results, such as point estimates, upper limits, lower limits, p-values, Z-values, and effect sizes are described in Fig. 2. It should be noted that some articles are observed in more than one row because they were reported in the results at least twice.
Forest plot.
Tests of homogeneity
To analyze the homogeneity of the studies, Cochran’s Q, τ-squared, and I-squared tests were applied (Borenstein et al., Reference Borenstein, Hedges and Rothstein2009). Then, two hypotheses were considered:
Null hypothesis: There is no significant difference between the obtained effect sizes.
Hypothesis (I): There is a significant difference between the obtained effect sizes.
The null hypothesis indicated a lack of significance for the obtained effect sizes, while the other hypothesis implied a significant difference between the obtained effect sizes. Since the level of significance at the 95% confidence interval was lower than the error margin (5%), H0 was rejected and H(I) was confirmed. In other words, a significant difference existed between the obtained effect sizes which indicated the heterogeneity of the obtained effect sizes. In addition, the I-squared and τ-squared values that are convenient criteria to determine the homogeneity or heterogeneity of data pinpointed a high degree of data heterogeneity (Table 1).
The results of the meta-analysis
* p-value = .000, **Place of looking for missing studies, ***Number of missing studies that would bring p-value to alpha.
The meta-analysis results
In meta-analysis, the magnitude of the relationship between two variables is measured by effect size (Kelley & Preacher, Reference Kelley and Preacher2012). The meta-analysis demonstrated a significant relationship between innovation and organizational performance. The random-effects and fixed-effects models are illustrated in Table 1. The appropriate model is developed by identifying the homogeneity or heterogeneity of the data. Furthermore, the results of Q and 
${{\text{I}}^2}$ Cochrane tests, which are related to data homogeneity, were used to select fixed-effects or random-effects models. The heterogeneity observed between the studies (Q test: p = .000 and 
${{\text{I}}^2}$ = 94.316%) indicates that the data were random; therefore, the random-effects model was employed. The random-effects model demonstrates that the research studies were conducted randomly (Borenstein, Hedges, Higgins, & Rothstein, Reference Borenstein, Hedges, Higgins and Rothstein2010, Reference Borenstein, Hedges, Higgins and Rothstein2011).
The effect size value obtained for this relationship is 0.492 (95% CI = 0.462–0.522; p = .000). The finding supports H1 supposing that innovation relates positively with organizational performance. As noted, Cohen (Reference Cohen1992) determined the effect sizes equal to 0.1 as small; 0.3 as a medium; and 0.5 as large effect sizes (Cohen, Reference Cohen1992; Jing, Reference Jing2018). The obtained effect size is between 0.3 and 0.5; therefore, it is concluded that the effect size is medium. Moreover, the obtained p-value is less than 0.05, which demonstrates the significance of this relationship with a medium effect size. The correlation coefficients and the combined results are represented by a forest plot (Borenstein et al., Reference Borenstein, Hedges, Higgins and Rothstein2011). The forest plot of the random-effects model is also illustrated in Fig. 2.
Analysis biases
Selection bias
Selection bias occurs when the researcher removes or selects studies based on personal judgment (Egger & Smith, Reference Egger and Smith1998). As stated in the ‘Research Methodology’ section, an appropriate and specific process and protocol have been applied to prevent selection bias.
Citation bias
Citation bias occurs when researchers use just one database or journal to receive articles from. This bias also occurs when researchers try to cite the articles with more significant or larger effect sizes and ignore the articles with non-significant or smaller effect sizes (Jannot, Agoritsas, Gayet-Ageron, & Perneger, Reference Jannot, Agoritsas, Gayet-Ageron and Perneger2013). Four databases have been examined, and the magnitudes of the effect sizes reported in the articles have not been considered as a removing criterion in the protocol to prevent citation bias in the present study.
Linguistic bias
Language bias means that due to the lack of familiarity with other languages, the researcher may consider only one specific language and skip examining studies in different languages (Egger et al., Reference Egger, Smith and Phillips1997). It should be pointed out that the focus of some databases is on indexing English journals instance, more than 90% of the articles indexed in Scopus are English (Albarillo, Reference Albarillo2014). Thus, merely limiting the search of a structured review to a database such as Scopus will bring about linguistic bias. In the current study, three English databases and a Persian database (Magiran) were used to avoid linguistic bias.
Publication bias
Articles with small effect sizes are less likely to be published than those with large or medium effect sizes. Scholars commonly refuse to publish articles with small effect sizes; this occurrence is called publication bias (Egger et al., Reference Egger, Smith and Phillips1997). The symmetry or asymmetry in the studies depends on the p-value in both one-tailed and two-tailed states; if this value is less than .05, it means that publication bias has occurred in the research (Borenstein et al., Reference Borenstein, Hedges, Higgins and Rothstein2011). The null hypothesis in this test was the symmetry of the studies, confirmed at the 95% confidence interval, since the p-value was above .05 (Table 1, p-value (1-tailed) = .06346 & p-value (2-tailed) = .12692). Thus, based on the funnel plot as illustrated in Fig. 3, the variables of the study were accumulated at the top and no publication bias was detected.
Funnel plot of standard error by fisher’s z.
Fitting correction by Duval and Tweedie’s method
Duval and Tweedie’s trim and fill technique was introduced to evaluate and modify publication bias in small samples. The findings demonstrated that 30 articles had to be added to the right side of the funnel plot to make it symmetric. This confirms the existence of publication bias in this study. The findings obtained from Egger’s test confirmed the lack of any publication bias, though Duval and Tweedie’s method pointed to a contradictory result. This indicated a weakness of the method for justifying publication bias and proved to be inconvenient for the current study. Another shortcoming was that the method failed to prove the insignificance of the results when the missing studies are added. Thus, classic fail-safe N was applied to examine the reliability of the findings.
Rosenthal’s fail-safe N test
The ‘fail-safe N’ test demonstrates the number of required studies that, if added to the analyses, a statistically insignificant overall effect will be obtained (Borenstein et al., Reference Borenstein, Hedges, Higgins and Rothstein2011). Table 1 illustrates the results of the classic fail-safe N test. According to Table 1, the Rosenthal’s fail-safe N value is equal to 10,755. Accordingly, 10,755 studies had to be added to make the results insignificant. According to a study by Rosenberg (Reference Rosenberg2005), if the fail-safe N value in a study is higher than 500, the study will be highly validated (Rosenberg, Reference Rosenberg2005).
On the other hand, according to the formula presented by Rosenthal (Reference Rosenthal1986) if the fail-safe N value (N = 10755) is higher than 625, the validity will be approved (Rosenthal, Reference Rosenthal1986). Furthermore, based on the formula N/(5K + 10) > 1 where K (180) indicates the number of studies and N (10755) indicates the value of the classic fail-safe N, the lack of any publication bias and the accuracy and reliability of the results can be proven (Mullen, Muellerleile, & Bryant, Reference Mullen, Muellerleile and Bryant2001).
Moderator variables
A variable that can directly affect the direction or extent of the relationship between independent and dependent variables is called the moderator variable and it is noteworthy that the results of these effects are measurable (Borenstein et al., Reference Borenstein, Hedges, Higgins and Rothstein2011). In the context of meta-analysis, the Q test or Cochran’s Q test is used to assess heterogeneity among the effect sizes of different studies. The null hypothesis for the Q test is that all studies are evaluating the same effect, meaning there is no heterogeneity (Borenstein et al., Reference Borenstein, Hedges and Rothstein2009). If the Q statistic is large and the associated p-value is small (typically less than .05 or .1), this provides evidence to reject the null hypothesis, indicating that there is significant heterogeneity among the studies. This could suggest the presence of a moderator variable that is causing the differences in effect sizes among the studies (Cochran, Reference Cochran1950; Conover, Reference Conover1999).
According to the large Q-value and the small p-value (Q-value = 3149.31, p-value = .0001), the current study considered the country, continent, publication year, and innovation type as the hypothesized moderator variables (Table 2). The first and second moderating variables were the country and the continent in which the study was conducted. Therefore, each effect size was coded into 32 countries and 4 categories of continent (North United States, South United States, Europe, Asia, and Africa). Two papers were not considered in any group because they were from countries of different continents.
Categorical moderator analysis results
Q B = test of difference between moderator level, k = number of data, r = mean correlation, z = between group difference. (marginally) † p < .1,
* p < .05, **p < .01, ***p < .001.
The third moderator was year of publication. Finally, the variable of innovation type was examined as the fourth moderator variable. In this study, the types of innovation were not grouped in order to independently examine the impact of each type on firm performance, allowing for the investigation of moderators of innovation types on organizational performance. This approach enables a detailed analysis of the differential effects of various innovation types on firm performance, contributing to a more comprehensive understanding of the relationship between innovation and organizational outcomes. A total of 180 data examined the relationship between 47 different types of innovation and organizational performance. Innovation types with insufficient data points, especially less than three, were excluded from the analysis in the moderator analysis section. According to Table 2, the relationship between 13 types of innovation and organizational performance was investigated separately.
The examination of moderating variables in the study reveals that country, continent, publication year, and type of innovation all have a moderating effect on the overall effect size, with varying degrees of intensity. Table 2 lists the major information about each country and the results confirm the country (
${{\text{Q}}_{\text{B}}} = $ 351.033, p = .0000) and continent (
${{\text{Q}}_{\text{B}}} = $ 54.057, p = .0000) as a moderator variable. The finding supports H2 and H3 supposing that continent and country is a moderator variable for the relationship of innovation and organizational performance. For interpretation purposes, only those countries that contributed five or more effect sizes will be discussed. The results illustrated that Indonesia (r = 0.511, p = .000), Malaysia (r = 0.549, p = .000), Iran (r = 0.526, p = .000), and South Korea (r = 0.515, p = .000) moderate the overall effect size with a high effect intensity. Furthermore, Brazil (r = 0479, p = .000), China (r = 0.434, p = .000), UAE (r = 0.360, p = .000), Pakistan (r = 0.490, p = .000), Spain (r = 0.417, p = .000), Taiwan (r = 0.419, p = .000) moderate the overall effect size with low effect intensity.
The continents of Asia (r = 0.494, p = .000) and Europe (r = 0.513, p = .000) moderate the overall effect size with a high effect intensity, whereas the continents of Africa (r = 0.298, p = .000) and South United States (r = 0.479, p = .000) moderate the overall effect size with low impact intensity.
The year of publication was identified as a marginal moderating variable and H4 was supported (
${{\text{Q}}_{\text{B}}} = $15.652, p = .0745). According to the findings from the ‘Test of difference between moderator level’ it can be observed that the categorization of articles into subgroups based on publication year demonstrates a marginally moderating effect (
${Q_B}$=15.652, p < .1) on the association between innovation and organizational performance. The analysis of Table 2 and Fig. 4 reveals that among the articles published in the past decade, those from 2014 exhibited the most substantial effect size (r = 0.704), indicating a noteworthy contribution that strengthens the overall impact (r = 0.492).
Correlation between innovation and OP from 2012 to 2022 (January).
As can be seen in Table 2 and Fig. 4, the papers published in 2013 (r = 0.647, p = .000) and 2014 (r = 0.704, p = .000) had the highest effect size in the last 10 years, while the papers recently published in 2019 (r = 0.442, p = .000), this effect size reached its lowest level in the last 10 years.
Finally, the type of innovation was identified as a moderating variable in this study and the results confirmed H5 (
${{\text{Q}}_{\text{B}}} = $35.418, p = .0012). For interpretation purposes, only those types that contributed seven or more were reported will be discussed. Table 2 provides the information about each innovation type. The results illustrated that organizational innovation (r = 0.527, p = .000) moderate the overall effect size with a high effect intensity, while the Product/service innovation (r = 0.475, p = .000), Process innovation (r = 0.417, p = .000), Open innovation (r = 0.423, p = .000), Administrative/Management innovation (r = 0.477, p = .000), and Green/Environmental innovation (r = 0.398, p = .000) moderate the overall effect size with a low effect intensity.
Discussion and implications
Meta-analyses provide a more comprehensive view of the impact of diverse variables by integrating the results of different studies conducted on various samples. This study examines the relationship between innovation and organizational performance. This study offers a unique approach compared to other works in the field. Unlike previous research that categorized various types of innovation, this investigation focused on the independent examination of each type of innovation. By adopting this approach, the analysis aimed to: (1) Maintain the separation and distinctiveness of the different types of innovation and (2) Avoid potential confounding effects that could arise from integrating the types of innovation. This methodological choice emphasizes the importance of considering the specific types of innovation in the context of organizational performance and highlights the need for further research to better understand the relationship between innovation and organizational outcomes.
Significant heterogeneity was detected in the data after conducting the necessary examinations, and consequently, a ‘random-effects model’ was applied to continue the work. Finally, the relationship between innovation and organizational performance was calculated to be significant (sig ≤ 0.01) and medium (r = 0.492) according to Cohen’s criterion.
The findings of this meta-analysis were illustrated to be in line with some of the investigated studies. The findings by Hang et al. (Reference Huang, Wu, Lu and Chia Lin2016) demonstrated that innovation as a key component of knowledge management increasingly determines the competitive advantages of companies. In addition, they argued that a significant relationship exists between innovation and organizational performance (with a correlation coefficient of 0.492). The results of this research are consistent with the results of Bowen’s research conducted in 2010. Bowen et al. evaluated the relationship between performance and innovation based on the actual temporal sequence and revealed this relationship is significantly positive (r = 0.16), which this effect size is increased in this research. (Bowen et al., Reference Bowen, Rostami and Steel2010). Widjaja, Wihardja Sumintapura, and Yani (Reference Widjaja, Wihardja Sumintapura and Yani2020) demonstrate that there is not always a significant relationship between organizational performance and innovation, emphasizing the importance of additional research and the consideration of contextual factors (Widjaja, Sumintapura, and Yani Reference Widjaja, Wihardja Sumintapura and Yani2020).
Inconsistencies in the results of previous research can make it difficult for researchers to draw conclusions. Therefore, one of the academic implications of this systemic review is that it addresses this research gap and provides a single response to the relationship between innovation and organizational performance. Furthermore, researchers can compare the results of their research on the relationship between innovation and organizational performance with the results of the present study to exert further analysis. Considering the overall integrated findings associated with this relationship can be useful in developing theories as well as creating new frameworks and models for future research agendas.
Different types of biases including selection, citation, linguistic, and publication biases were investigated. Based on the obtained results, the lack of any publication bias and the accuracy and reliability of the results were proven.
In the moderator identification section, four moderator variables of country, continent, year of publication, and types of innovation were found. The results demonstrated that from among the countries with five or more data reports, the effect size of the relationship between innovation and organizational performance in Iran (r = 0.526, k = 39), Indonesia (r = 0.511, k = 8) and South Korea (r = 0.515, k = 5) was higher, while the lowest effect size belonged to the UAE (r = 0.360, k = 8). From among the continents, Europe (r = 0.513, k = 28) had the largest effect size between innovation and organizational performance, while the lowest effect size belongs to Africa (r = 0.298, k = 5).
Moreover, the results of moderating analysis demonstrated that there is a less significant relationship between innovation and organizational performance in Spain (r = 0.4173, k = 12, p = .0000) than in other European countries (r = 0.585, k = 17, p = .0000). Thus, Spain was identified as a moderating variable in Europe (
${{\text{Q}}_{\text{B}}} = $6.1006, p = .0135).
As illustrated in Fig. 4, the impact of innovation on organizational performance has decreased in recent years; therefore, this amount reached below the overall impact factor in the last 4 years. According to the tradeline (Fig. 4), whereas there are significant differences between the annual results obtained in the first half of the graph, in the second half, a significant convergence between the results in recent compatible years is observed. Therefore, the results and the viewpoints of researchers have been made compatible with each other in recent years. The results demonstrated that innovation performance strengthens the overall effect size among various types of innovation and this illustrate that innovation performance plays a significant role in enhancing the relationship between innovation and firm performance. This is consistent with prior research indicating the positive impact of innovation on organizational outcomes, and it underscores the importance of effectively managing and leveraging innovation activities to achieve superior performance.
As revealed by the review conducted by the publication year moderator, the relationship between innovation and organizational performance in research conducted before 2016 exhibited considerable fluctuations. However, recent years have seen a dearth of comprehensive studies examining this relationship across diverse work domains and various types of innovation. Consequently, conducting an all-encompassing investigation became imperative. Notably, this research introduces several innovative aspects, including the consideration of moderators such as the publication year, country, continent, and type of innovation.
Interestingly, past meta-analysis articles, which typically encompassed research up to 2012 and even as far back as 1989, reported a weaker relationship compared to the findings of this study (Bowen et al., Reference Bowen, Rostami and Steel2010; Feng et al., Reference Feng, Ma and Jiang2020). The divergence could be attributed to the heightened significance of innovation, rapid technological advancements, and the emergence of new project management methodologies that foster creativity – such as the agile method – in recent years. The experimental findings of the current study demonstrated several applicable outcomes for the managers. First, innovation stimulates companies to actively seek external resources, cooperate with external partners, improve the productivity of innovation, and shorten the time to respond to the market. On the other hand, innovation influences customers’ knowledge of the quality of trademarks and indirectly improves customer loyalty. Though companies usually consider innovation as a positive instrument for organizations, the relationship between innovation and performance is still an open question (Liao & Rice, Reference Liao and Rice2010). Thus, managers should pay specific attention to innovation, encourage customer participation, and merge internal and external resources. Second, the relationship between innovation and performance is influenced by the attitudes of a company toward risks.
Innovation performance in areas with weaker uncertainty avoidance is better than the areas in which uncertainty avoidance is dominant. Therefore, managers should pay enough attention to lay the grounds for accepting modern innovations and encourage their employees to move away from their ‘comfort zone’ to innovate with audacity. In general, an organization that can present new ideas will be able to adopt changes well and act as an agent of change. Thus, organizations should adopt creativity and innovation. Otherwise, they will be forced out of the competition.
Limitations and suggestions
The current study suffers from both temporal and database limitations. The search process was conducted in four databases – Scopus, ASCE, ScienceDirect, and Magiran (2012–2022). In addition, the current study attempted to counteract the publication bias by including unpublished studies. However, access to such experimental studies was constrained. Nevertheless, it is not expected that a strong normative bias exists on the side of reviewers and editors concerning the publication of non-significant or the establishment of negative relationships between innovation and performance. The addition of such unpublished studies will certainly establish a stronger context for the empirical generalization of their impacts.
Different tests concerning the assessment of effect size, publication bias, and homogeneity indicated the accuracy of the findings. Furthermore, four moderator variables that are among the most important findings were found in this meta-analysis. Thus, detecting other moderator variables can be considered and future researchers may attempt to detect them in their studies. The investigations conducted in the current study led to the discovery of different aspects of innovations. Performing a systematic review for each aspect can be useful. Finally, it should be stated that the current study attempted to investigate the relationship between innovation and organizational performance by the integration of effect sizes. However, this is not enough, and the same study should be replicated using other databases, periods, and instruments, and preferably the qualitative methodologies in order to compare the obtained results with the findings of the current study.
Conclusion
The current meta-analysis was conducted to determine the combined effect size of the relationship between innovation and organizational performance. English and Persian articles published from 2012 to 2022 were explored in four databases. One hundred and forty three articles were selected for analysis by CMA2, and it was illustrated that the total combined effect sizes of innovation on organizational performance for fixed- and random-effects models are 0.481 and 0.492, respectively. The results of Cochran’s Q test demonstrated that the data are heterogeneous. Thus, the average effect size obtained from the random-effects model (0.492) was considered as the final approach. Cohen’s coefficient was demonstrated that the strength of the relationship between innovation and organizational performance is medium considering the approach taken by the random-effects model.
Different types of biases including selection, citation, linguistic, and publication biases were investigated. Based on the obtained results, the absence of any publication bias and the accuracy and reliability of the results were proven. The upper and lower limits expressed in this study have a wide range. This range demonstrates that the relationship between innovation and organizational performance may vary depending on the moderating variables. The analysis was conducted to find the moderator variable.
Finally, country, continent, year of publication, and innovation type were identified as the moderator variables. Based on the results, researchers can conduct more extensive studies on these modulatory variables. Furthermore, the results of this study can enable managers of different departments in different industries to have a more comprehensive view of the issue, and a better planning for improving their organization performance through innovation.
The results illustrated a significant difference in the relationship between innovation and organizational performance in different countries and continents. Therefore, researchers can continue the research with a new range of research to investigate the reasons for this discrepancy. In addition, the results demonstrated a declining trend in the relationship between innovation and organizational performance over the past 10 years. Researchers are encouraged to investigate the reason for this declining trend and identify new factors that have a growing impact on organizational performance. Investigating the mediating and moderating role of innovation in organizational performance could be a valuable objective for future meta-analysis studies.
Competing interest
No conflict of interest exists. We wish to confirm that there are no known conflicts of interest associated with this publication and there has been no significant financial support for this work that could have influenced its outcome.
