Emil Israel (p), Eyal Salinger

Innovation is a diffusional process, according to much research in geography. In this process, saturation in economic cores of national and international markets increases the probability of an innovation trickling out and affecting other regions. Economic cores provide the best prospects for innovation in comparison to lagging regions. Economic cores, which are usually metropolitan regions, are characterized by an innovative ‘milieu’ with ample and appropriate resources to stimulate innovation, particularly technological advances, that stems from R&D investments and introduction of new products. Conversely, lagging regions lacking significant urban agglomerations suffer from structural inferiority due to, among other things, lack of an appropriate milieu. These shortages include lack of a skilled labor force, infrastructural inferiority, absence of training facilities, and social and organizational culture that does not encourage (technological) innovation. To contend with this inferiority, contemporary regional policies emphasize new aspects of innovation. These endeavors advance non-technological forms of innovation, which do not necessarily require significant investments in R&D. It is believed that such a shift could stimulate regional innovation, and thus promote economic growth of lagging regions. This study explores different forms of regional innovation by looking at an empirical case study that concentrates on a comparison between a core region and periphery in Israel. The current inquiry utilizes the innovation surveys of the Central Bureau of Statistics of Israel conducted in the years 2006-2008 and 2010-2012. These surveys sampled more than 4,000 firms in different economic sectors, within different parts of the country, enabling the distinction between technological and non-technological innovation outcomes (for example innovative products and processes) that these firms introduced in different geographical areas. The study uses logistic regression models to test for the probability that firms introduce an innovative output, as function of their location, investments in R&D, economic sectors to which they belong to, technological characteristics, as well as their size, degree of multi-nationality and their seniority in the market. The results show that overall, differences in the probability of a firm to technologically innovate are not necessarily the outcome of their spatial location. However, the location of firms in the periphery decreases the probability that a firm will non-technologically innovate in comparison to location in Israel's main metropolitan region. The findings also indicate that firms in the periphery will conduct more non-technological innovation if they were in core regions. The results have policy implications for the promotion of regional economic growth.

Carlos Llano (p), John Forth, Santiago Pérez-Balsalobre, Ana Rincón-Aznar

The evaluation of the Single European Market requires a better knowledge of the level of integration both between and within the EU countries. While some institutions are pushing towards greater integration between the EU countries themselves, some others may be introducing -on purpose or collaterally- additional barriers to economic interaction within a given EU country. Several reports have denunced a high level of market fragmentation within Spain, mainly due to legal measures adopted by regions and municipalities. With the objective of deepening the integration of the internal market Spain approved the Law 20/2013 of 9 December on the Guarantee of Market Unity, which has been subject to critizism, mainly after the Spanish Constitutional court declared null one of the key provisions in June 2017. In this context, the aim of this paper is to study whether regional borders influence the pattern of commercial transactions across Spain. For this purpose, we depart from the border-effect literature, which offers a valuable basis for discussion, since it allows us to quantify the higher intensity of trade within a certain spatial unit in comparison with the intensity in other markets. Our empirical analysis covers a large time period (1995-2015), considering product specific flows between the Spanish regions (Nuts 2) and between them and the main countries. Our results suggest the presence of heterogeneity for the home-bias both at the spatial and sectoral level, finding in general high levels of integration, but also exceptions for certain regions and products.

Daniela Bragoli (p), Camilla Ferretti, Piero Ganugi, Giovanni Marseguerra, Davide Mezzogori, Francesco Zammori

Following the seminal work by Altman (1968) and a more recent contribution by Barboza, Kimura and Altman (2017), the aim of this study is to provide a model for the Italian economy that correctly classifies solvent and insolvent firms one year prior to the event, using AIDA Bureau van Dijk dataset over the period 2007-2016.
By applying a full battery of bankruptcy forecasting models, which combine more traditional models such as logistic regression with more sophisticated techniques based on machine learning, the aim of this work is not only to provide accurate forecasts, but also to learn what variables are important predictors of firm bankruptcy in Italy.
For this purpose we add to the financial indicators used in the literature (excluding market based indicators since only a very small number of Italian firms is listed) a set of new variables that describe the industrial structure of the firm, including district membership, but also a measure of mark up, the level of vertical integration, sectorial and regional dimensions.

Results show that the model for the Italian economy provides accuracy indicators that are in line with those in the literature (Barboza, Kimura and Altman, 2017). We also find that, in terms of accuracy, computational methods perform relatively better than the logit. In addition, thanks to the logit regression, we are able to learn the determinants of the probability of default for Italian firms. Interestingly both the financial ratios and the industrial structure variables seem relevant in terms of influencing the probability of default. In particular industrial district membership reduces the probability of bankruptcy.

From a methodological point of view, we can distinguish two parts. In the first part, we conduct the bankruptcy forecasting exercise. We thus divide the dataset into a training set in which we estimate several models and a testing set in which we validate which model is able to correctly classify solvent and insolvent firms. We consider several models, the more traditional logistic model and more state of the art machine learning techniques (such as random forest, artificial neural networks). We report model performance tests for each model. For robustness, we also report cross validation results.
In the second part of the paper, we focus on the logistic regression of the dependent dummy variable (bankruptcy) on the independent variables, which include financial ratios, district membership and other industrial variables, sector and regional dummies. We report the coefficients and p-values.

Ferran Vendrell-herrero, Christian Darko , Yancy Vaillant (p)

For the last three decades cluster policy has been an extremely resilient policy to enhance regional competitiveness. Whilst there are various cases of success (i.e. Basque Country) there is still the need to develop policy instruments that enable regulators and cluster managers to better respond to social demands and economic challenges. This is particularly important in business environments that fall behind in terms of institutional development and public funding support for businesses. We argue that this is precisely the conditions faced of manufacturing firms in some African countries.
Public budget constraints in many African countries limit their ability to provide the necessary level of monetary support needed to implement hard industrial policies to support the competitiveness of African firms. Other mechanisms therefore such as soft policies can become important to support and facilitate local firm development. In this vein the creation of cluster zones has been advocated as a way to stimulate local firms to compete and ultimately increase their productivity. Cluster membership may be particularly important for SMEs as they lack the scale and resources of large MNEs to invest in innovation projects. The advantages to clustered firms are associated with connectedness although the conditions for these advantages to emerge are predicated on institutional support and social exchange. Based on these arguments we hypothesize that cluster membership increase productivity in African manufacturing firms; and that this effect is larger when the country has devoted intentional resources in order to implement this industrial strategy.
In order to respond to this question the authors draw on the World Bank Enterprise Survey. The cross-sectional surveys conducted in nine African Countries during 2010 contain 1,111 African SMEs in various manufacturing industries. The sample includes the question on whether the firm belongs to a cluster association. In the sample 28% of the firms are members of cluster associations. As a measure of firm productivity we use labour productivity. The treatment effect for both, the full sample and matched sub-sample constructed using nearest neighbour propensity matching score, indicate that cluster associates are more productive than non-cluster firms. We implement a series of counterfactual exercises to understand how non-cluster firms would benefit from joining a cluster association and conclude that the benefit is stronger when the government develops an intentional industrial policy towards the use of cluster zones. This result opens an interesting avenue for further research.

Gabriela Barrere (p), Andres Jung , Diego Karsaclián

The purpose of this study is to shed some light into the scarcely explored role of destination markets in the relationship between innovation and exports for developing economies. In this way we intend to contribute to the debate on a key issue for the design and implementation of innovation and internationalization policies in developing countries.
Innovation has turned into one of the most important features of the political agenda of both developed and developing economies. New products and new processes are the way firms access external markets and participate in global value chains. Developing economies have a more concentrated and less sophisticated structure of production and exports than developed countries. This lack of ‘complexity’ hampers their possibility of achieving sustainable high growth rates, unless they enter a path of diversification. This path of diversification of the productive and export structure requires innovation at the firm level. The key question is if innovation is necessary to be more efficient and, in this way, access export markets, or if it is the participation in export markets the way firms access to new knowledge that enable further innovation. These two approaches, largely discussed in the literature and non-mutually exclusive, have been referred to as the self-selection hypothesis and the learning-by-exporting hypothesis, respectively. The former stresses that more productive firms are able to overcome sunk costs of entering and prevailing in international markets before starting the export activity. The latter suggests the idea of a learning process, originated in the experience of participating in international markets, that would impact on innovation and productivity growth at the firm level.
Some studies have shown mixed results, with the prevailing effect depending on the characteristics of the firm or finding a bidirectional relation and a certain level of endogeneity between these variables. For developing economies, findings are not yet conclusive and, particularly for Latin-America, there are few empirical studies.
Our main hypothesis is that, for developing countries, learning-by-exporting effects are present mainly when the destination market is a developed economy, but that innovation driven self-selection is dominant when the export destination market is a developing economy.
We apply a bivariate probit estimation, fitting a maximum likelihood two-equation probit model, to three waves of survey data (2007-2009, 2010-2012 and 2013-2015) for 1321 Uruguayan firms, from manufacturing and services sectors. The data source is the Innovation Annual Survey for Uruguay, from the Agencia Nacional de Investigación e Innovación (ANII).