TRANSITION REPORT 2014 Innovation in Transition

Does firm-level innovation pay off?

Our analysis now turns to the relationship between innovation and the productivity of firms. Policy-makers and researchers widely acknowledge that innovation is essential for increasing productivity.8 However, while a positive correlation between product innovation and firms’ performance has been established for European firms, evidence for developing countries has been mixed.9 Similar studies exist only for a subset of transition countries. Indeed, for many of them, the data required for such analysis have not existed until now.

A simple comparison of the average labour productivity of innovative and non-innovative firms does not point to a strong relationship between innovation and productivity. Innovative firms have higher average productivity in less than half of all countries. Differences between innovative and non-innovative firms also depend on the type of innovation. Only in Jordan are innovative firms significantly more productive than non-innovative firms across all types of innovation (see Table 2.1).

Read more

There may be reasons why the correlation observed between innovation and productivity is weaker than the true underlying impact that innovation has on productivity. For example, if poorly performing firms find themselves under greater pressure to innovate, innovation may appear to be linked to poor short-term performance, despite improving firms’ productivity in the longer run.

In order to deal with such issues appropriately, we need a more comprehensive analysis of the relationship between innovation and firms’ productivity that accounts for factors that may affect both firms’ productivity and the decision to innovate. To this end, this chapter uses a well-known model devised by Crépon, Duguet and Mairesse (known as the “CDM model” – see Chart 2.1.1) that links R&D, innovation and labour productivity.10 The model controls for other factors that can affect R&D, innovation and labour productivity, such as a firm’s size and age, the skills of the workforce, the level of competition and the type of industry (see Box 2.1 for details).

Once these factors are taken into account, the impact that innovation has on productivity becomes stronger.11 Product innovation is associated with a 43 per cent increase in labour productivity, and this effect has a high degree of statistical significance. This suggests that a firm with median labour productivity would move from the 50th to the 60th percentile of the labour productivity distribution after introducing a new product. Labour productivity also benefits from the implementation of process innovations. Although this effect is smaller (with the introduction of new processes being associated with a 20 per cent increase in labour productivity), it is also statistically significant. A firm with median labour productivity would move from the 50th to the 55th percentile of the labour productivity distribution after introducing a new process. These effects are somewhat stronger than those found for developed economies, but they are comparable to those observed in developing economies.12

Interestingly, the increase in labour productivity is smaller for firms that engage in product and process innovation simultaneously than it is for those that engage exclusively in either product or process innovation. This can be explained by the fact that simultaneous product and process innovation is more complex and takes longer to be fully reflected in increased labour productivity, while BEEPS data only enable us to look at the short-term impact of new products and processes.

The estimated effects are stronger when using self-reported measures of innovation than when using cleaned measures. In the case of product innovation, the estimated improvement in productivity is 69 per cent when a self-reported measure of innovation is used, compared with a 43 per cent improvement when using a cleaned measure. This could be because almost a quarter of all self-reported product innovations and 11 per cent of all self-reported process innovations were in fact either organisational or marketing innovations (see Box 1.1), which nevertheless result in increased turnover per worker. Indeed, the increase in labour productivity associated with self-reported organisational and/or marketing innovation is estimated at 67 per cent.13 Organisational and marketing innovations are probably less risky and costly for firms than technological innovations and, given these high productivity yields, it is perhaps surprising that less than a third of all BEEPS firms engage in either. This could be due to a lack of information on new organisational and marketing methods, scepticism regarding their effectiveness or resistance to change within organisations.14

Firms that innovate are more productive in less than half of all transition countries
Level of significance

Type of innovative activity





Jordan, Moldova, Romania, Russia

Armenia, Croatia,
FYR Macedonia


Product and process


Kyrgyz Rep., Moldova, Mongolia

Armenia, FYR Macedonia, Uzbekistan

Organisational and marketing (self-reported)

Belarus, Jordan, Latvia, Russia, Slovenia

Kyrgyz Rep., Lithuania, Mongolia, Romania, Tajikistan, Turkey

Kazakhstan, Montenegro

Product and process (cleaned)


Jordan, Moldova, Mongolia, Ukraine


Source: BEEPS V, MENA ES and authors’ calculations.
Note: There are no BEEPS firms engaged in research and development (R&D) in Azerbaijan. Cleaned data on product and process innovation were not available for the Slovak Republic, Tajikistan or Turkey at the time of writing.


Table 2.2. The impact of innovation on labour productivity depends on the type of innovation
  Associated impact on firm-level productivity 
  (1) (2)
Type of innovation Cleaned Self-reported

Product innovation





Process innovation





Product or process innovation





Non-technical innovation
(marketing or organisational)



Source: BEEPS V, MENA ES and authors’ calculations.
Note: This table reports regression coefficients for the occurrence of innovation at firm level, reflecting the impact on the dependent variable firm-level productivity, which is measured as turnover (in US dollars) per employee in log terms. The results are obtained by estimating a three-stage CDM model by asymptotic least squares (ALS), where productivity is linked to innovation, and innovation, in turn, is related to investment in R&D. For a detailed description and the set of control variables included, refer to Box 2.1. Standard errors are reported in parentheses below the coefficient. ***, ** and * denote statistical significance at the 1, 5 and 10 per cent levels respectively.