![Sample our Economics, Finance,Business & Industry journals, sign in here to start your access, latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5931/TOC-Subject-Sample-2021-Economics-Finance-Business-Industry.jpg"})
ABSTRACT
This is the first study to analyse the contribution of context to firms’ perception of innovation barriers in a single country. Using the Ecuadorian Innovation Survey and multilevel logit models, we study whether the geographical location of Ecuadorian firms makes them more likely to assess three financial, five knowledge and two market barriers as relevant factors hindering their innovation activities. Our results indicate that location in one of Ecuador’s 24 regions has only a subtle effect on perception of barriers. After controlling for internal and sectoral characteristics of firms in each region, we find that only 2–6% of the dispersion observed for whether a barrier is perceived as relevant is due to regional differences. For financial and knowledge barriers, half of that small geographical component disappears when the model includes regional population density. Based on the latter result, we argue that urban economics arguments can explain the spatial distribution of firms’ perception of innovation barriers in this small developing country. Our results provide a critical reflection to advance the current research agenda on contextual factors affecting innovation.
Acknowledgements
The authors wish to thank the professors in the Department of Urban Rural and Territorial Studies of FLACSO Ecuador for their valuable suggestions.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 D’Este et al. (Citation2012) included regional dummies in their multivariate probit model on the determinants of firms’ perception of innovation variables in the UK. Hölzl and Janger (Citation2014) examined perception of barriers across European countries using dummies that group countries according to their distance from the technological frontier.
2 Only Khiari and ben Rejeb (Citation2018) have used multilevel models with data on firms’ perception of innovation barriers. Their study for Tunisia, however, is about the effects of perceived financial barriers on the intensity of innovation. These authors argue that the impact of financial barriers differs between regions due to each region’s economic particularity and infrastructure.
3 Ecuador is divided into 24 s-order administrative units, called ‘provinces’, which constitute our location variable. Henceforth, we use the term ‘region’ to refer to these provinces.
4 Cuenca, the third largest city, accounts for just over 3% of Ecuador’s total population.
5 Note that some studies include an additional aggregated category: regulation-related barriers.
6 Financial barriers include (1) high innovation costs, (2) lack of internal funds and (3) lack of external funds. Knowledge barriers include (4) lack of qualified personnel in the firm, (5) lack of qualified personnel in the country, (6) lack of technological knowledge, (7) lack of market knowledge and (8) difficulty finding cooperation partners. Market barriers include (9) markets dominated by established firms and (10) uncertainty of demand for innovations.
7 Our empirical analysis, however, focuses on the barriers that emerge after firms engage in the innovation process, as the ENAI only provides information on perceived barriers for innovative firms.
8 ‘Lack of market knowledge’ refers to issues such as high entry cots, technology, product- and service-requirements or poor understanding of diffusion timing or alterations in relative factor prices after innovation (Caiazza Citation2016).
10 Due to a lag between sample selection and data collection, 30 firms in our final sample reported fewer than 10 employees in each of the 3 years covered by the survey. We retained these firms in our sample.
11 of the ENAI’s methodological document shows the number of firms included in the sample by province: https://www.ecuadorencifras.gob.ec/documentos/web-inec/Estadisticas_Economicas/Ciencia_Tecnologia-ACTI/2012-2014/Innovacion/MetodologIa%20INN%202015.pdf.
12 Of the firms in our sample, 78.1% have a unique establishment. Repeating the analysis presented in and for these firms produces very similar results, which are available upon request.
13 We also lost two observations because of missing data on employees’ qualification level.
14 In the ENAI, firms evaluate barriers according to the following scale: high, medium, low, non- relevant.
15 Our sample includes only two firms in Galapagos, three in Zamora Chinchipe and five in Morona Santiago. The mean number of firms for the sample in the other 21 regions is 124, and the median is 54. As discussed in the next section, multilevel modelling is suitable for this type of unbalanced sample.
16 According to the 2015 Ecuadorian Innovation Survey, 64% of innovative firms stress that financial barriers hinder their innovation activities, while 57% point to knowledge barriers and 54% to market barriers.
17 We have used data from the Ecuadorian National Institute for Statistics and Censuses and our own calculations based on GADM spatial country data on administrative areas.
18 Another way to examine this issue would be to analyse product and process innovators separately (Hölzl and Janger Citation2014).
19 Although studies of innovation barriers tend to focus only on service and manufacturing firms, we included firms in the mining and quarrying sector because they were reporting innovation barriers as factors hindering their innovation activities. As indicates, 3% of firms in the final sample belong to this sector.
20 Some previous literature on innovation (De Fuentes, Santiago, and Temel Citation2020) and innovation barriers (Santiago et al. Citation2017) has focused on sectoral heterogeneity. Our sectoral dummy variables capture average sectoral effects on perception of innovation barriers for firms in Ecuador. Our paper focuses on geographical contextual effects, and the sample size does not permit proper distinction of sectoral effects by region.
21 Detailed description of this model can be found in Srholec (Citation2010), López-Bazo and Motellón (Citation2018) and Bruna and Fernández-Sastre (Citation2021). See also the webpage of the Centre for Multilevel Modelling, University of Bristol.
22 The ‘odds’ are the ratio of probability of perceiving a barrier to probability of not perceiving it , and they can vary from zero to infinity. The natural logarithm transforms the set of positive real numbers into the whole real line. Estimates in show results for log odds, whose sign provides an intuitive interpretation of probability of perceiving innovation barriers.
23 Note that the original data for the dependent variables is binary, 0 or 1. The variance of the level-one residual error is assumed to be of standard logistic distribution: . Including individual or regional variables in the model can thus change only the estimated variance of the regional random effect. Still, the ICCs of the multilevel logit models may be used to compare the general magnitude of regional effects under different modelling equations.
24 The random effects are precision-weighted residuals termed ‘posterior residuals’, ‘empirical Bayes estimates’ or ‘shrunken residuals.’ They are estimated by considering the number of observations in each region and dispersion of the data within and between regions.
25 Although regional influence is modest, the knowledge barrier ‘lack of market knowledge’ exerts greater influence. The barrier with the lowest regional influence is ‘existence of a market dominated by established firms. It is likely that most firms compete at national level, such that competitors are the same regardless of the region in which they are located.
26 Our one-country study does not permit us to assess country heterogeneity related to national innovation systems. Our data from a national innovation survey also prevent us from studying contextual effects at municipal level. Moreover, urban, regional, and national effects could overlap and interrelate though different feedback mechanisms, bottlenecks, and synergies.
27 We estimated models including alternative regional variables that might capture the presence of large cities: Log gross value added, Log total population and Log urban population. The results are like those of Log population density and are available upon request. Conversely, Log gross value added per capita is not statistically significant.