Process innovations and firm productivity growth

Following the classification in the survey used in this work (see Sect. 2), for the empirical analysis we classify as large firms those with more than 200 employees and as small firms those between 10 and 200 employees.

This methodology has been previously used in this literature in Máñez et al. (2005), to analyse the relationship between R&D activities and productivity. In related literature, it is also used by Delgado et al. (2002) and Girma et al. (2004) to analyse the relationship between firm productivity and exports.

See http://​www.​funep.​es/​esee/​ing/​i_​esee.​asp for further details.

The information in the ESEE about these informal activities is collected on a 4-year basis.

Shane (2001) argues that one of the important attributes for one invention is its radicalness.

However, Audretsch and Vivarelli (1996) find that whereas large firms have a comparative advantage for innovation at exploiting knowledge created through R&D investments in their own labs, small ones comparative advantage stems from exploiting capabilities outside the firm, such as universities and the human capital characterizing a given region.

According to Hewitt-Dundas (2006), small firms are more likely to face capability constraints (such as human or organizational capabilities) to innovation than larger firms and, consequently, it is especially relevant for them to access to the complementary assets of external innovation partners (in terms of accessing specialist expertise, knowledge, etc.).

However, if we do not condition on undertaking R&D, the R&D intensity ratio is higher for large (1.43%) than for small firms (0.81%). These figures are not reported in Table 2.

Skilled labour has been calculated as the ratio of highly qualified workers (engineers, technical engineers and other university graduates) over the total number of workers.

Markard and Truffer (2006) find that only large firms can afford radical projects and that highly dedicated employees are identified to be a key driver for more radical innovation strategies.

Beneito (2006) uses the same database but for the period 1991–1996.

According to Dougherty and Hardy (1996), the combination of financial capability constraints with the risk in results associated to an innovation project make small firms more prone to choose low-risk innovation incremental projects, which have a more immediate reward.

Additionally, we have calculated the percentage of small and large firms that obtain patents and utility models. We get that 75.76% of small firms obtain patents (83.62% of large firms), independently of getting utility models simultaneously, and that 46.75% of small firms obtain utility models (42.93% of large firms), independently of getting patents simultaneously. This allows concluding that small firms are relatively more oriented to utility models with respect to patents than large firms.

Piva and Vivarelli (2005) find that Italian manufacturing firms implement intermediate/not-leading technologies mainly through non-R&D expenditures (informal innovation activities).

Acs and Audretsch (1990) find evidence that the innovative activities for small firms respond to a different technology and economic environment than that of large firms.

According to Baldwin (1997) small firms R&D activities are more occasional because they are generally aimed to solve production problems and to profit from market opportunities when they appear.

Cohen and Klepper (1996a, b) and Smolny (1998) argue that the larger the firm the greater the output over which it can apply the fruits (and spread the costs) of its R&D and hence the greater its returns from R&D. This possibility of exploiting innovations by incorporating them in their own output confers appropriability advantages to large firms. Cohen and Klepper (1996b) also argue that differences between large and small firms will be greater for process than for product innovations because the nexus between appropriability and size is closer for the former than for the latter (given that it is less likely for process innovations to be sold in disembodied form to other firms).

Should the productivity of process innovators previously to the introduction of the innovation be no lower than that of non-process innovators, process innovations will cause divergence. If it is lower, then process innovations will produce convergence between the productivity levels of process innovators and non-process innovators.

This is especially relevant for the case of R&D investments due to its sunk costs nature (Sutton 1991).

Some support for the self-selection hypothesis into R&D activities (formal part of innovation activities) of the most productive firms can be found, among others, in Hall 1990 (who uses a financial constraint argument); González and Jaumandreu 1998; González et al. 1999; and Máñez et al. 2005.

Another factor which may also contribute to a previously higher productivity level for process innovators is related to information prior to 1991, but we cannot control for it as we do not have any information before 1991. It could be the case that for some process innovators the difference in previous productivity comes from other innovations introduced before the starting year of the sample period.

We use a Gaussian kernel with a bandwidth parameter \( h = 0.9 \cdot A \cdot N^{{ - 1 \mathord{\left/ {\vphantom {1 5}} \right. \kern-\nulldelimiterspace} 5}} \), where A = minimum (standard deviation, interquartile rank/1.34) is estimated from the sample.

The TFP index itself allows controlling for industry factors that may also condition firms productivity (apart from introducing process innovations and the size of the firm). Another factor that could affect productivity is age. Although we do not control explicitly by age, we believe that the age effects on productivity have been, to a great extent, already captured by firm size. In the ESEE there is a strong relationship between age and size of firms.

See Handcock and Morris (1999) for the technical details about relative distributions.

The results in Calvo (2006) show that in Spain, over the period 1990–2000, small innovating firms have grown more than other type of firms.

We also report the yearly number of process and non-process innovators jointly with the empirical differences in the median value of TFP for these two groups of firms.

The cumulative yearly average growth rate is defined as: \( \ifmmode\expandafter\dot\else\expandafter\.\fi{z}_{{t_{f} - k_{f} ,t_{f} }} = {\left( {z_{{t_{f} }} - z_{{t_{f} - k_{f} }} } \right)}/k_{f} \).

The change in the definition for non-process innovators is required by the independence assumption in the KS tests.

When a first-time process innovator introduces a second innovation, its TFP growth from t + k to t + k + 1 is only calculated up to the previous period in which it introduces the second process innovation.

As shown in Table 8, to check for the robustness of our results we also test whether we reject the null of equality of the growth distributions for the subperiods (t + 2) − (t + 3) and (t + 3) − (t + 4). The results of the two-sided KS indicate that we cannot reject the null of equality for these subperiods.

Our results on Hypotheses 1 and 2 are compatible with the results of the analysis in Sect. 6.1 suggesting that for large firms the productivity level of process innovators in t is higher than that of non-process innovators in t only for 3 years of the sample (see Table 6). Whereas in Sect. 6.1 the comparison group is non-process innovators in t, in the analysis of hypothesis 1 and 2 are firms that never innovate. In Sect. 6.1, we could be comparing innovators in t with firms that do not innovate in t but were innovators in previous years and are still enjoying the effects on their productivity growth today. This problem is expected to be especially relevant for large firms as the time span of the effect of process innovations on productivity growth is longer, and a higher proportion of large firms are process innovators.

Máñez et al. (2005), using the same data to analyse the effects of undertaking R&D on productivity growth, obtain that undertaking R&D only fosters productivity growth for small firms (not for large firms) and for a very short period (from the period previous to the R&D investment to the period of the investment). The difference between Máñez et al. (2005) results and the results in this article points out that undertaking R&D does not ensure productivity growth if it does not result in a successful process innovation.

Later Schumpeterian thought has moved towards the idea of creative accumulation (Schumpeter 1942) that can be related to learning-by-doing and learn-to-learn processes.

Small firms may have a lower capability to attract and use skilled labour (Hewitt-Dundas 2006).

As Shane (2001) points out, radical technologies may destroy the capabilities of existing firms because they draw on new technical skills. If we link this argument with the relative deficiency on skill labour for small firms, we may expect small firms to be more reluctant towards radical technologies requiring different technical skills.

Levin et al. (1987) and Mansfield et al. (1981) find that whether major or incremental and whether patented or not, innovations are imitated within 1 year for many industries and within 3 years for most. Rents due to innovation are quickly dissipated and this makes even stronger the link between firm size (output) and appropriability advantages.