Do innovation and human capital explain the productivity gap between small and large firms?

Abstract

The economics of recombinant knowledge is a promising field of investigation. New technological systems emerge when strong cores of complementary knowledge consolidate and feed an array of coherent applications and implementations. However, diminishing returns to recombination eventually emerge, and the rates of growth of technological systems gradually decline. Empirical evidence based on analysis of the co-occurrence of technological classes within two or more patent applications, allows the identification and measurement of the dynamics of knowledge recombination. Our analysis focus on patent applications to the European Patent Office, in the period 1981-2003, and provides empirical evidence on the emergence of the new technological system based upon information and communication technologies (ICTs) and their wide scope of applications as the result of a process of knowledge recombination. The empirical investigation confirms that the recombination process has been more effective in countries characterized by higher levels of coherence and specialization of their knowledge space. Countries better able to master the recombinant generation of new technological knowledge have experienced higher rates of increase of national multifactor productivity growth.

Full text

Page 1

Research Institute of Applied Economics 2007 Working Papers 2007/16, 24 pages




Do innovation and human capital explain the
productivity gap between small and large firms?


Laia Castany♠, Enrique López-Bazo♣, Rosina Moreno♦
AQR-IREA
Universitat de Barcelona
Avda Diagonal 690, 08034 Barcelona (Spain)



Abstract:
Empirical evidence is compelling that large firms are more productive than
small firms. The hypothesis in this paper is that the productivity differences
between small and large firms are associated with two of the main determinants
of a firm’s performance: the human and technological capital that firms
incorporate. We suggest that the contribution of these factors in explaining the
productivity-size gap might not only be due to the fact that large firms make a
more extensive use of them, but also because large firms obtain higher returns
from their investment in human and technological capital. The evidence we
obtain for a comprehensive sample of Spanish manufacturing firms (1990-2002)
supports this hypothesis, which has important implications for the effectiveness
of policies designed to improve productivity in SMEs by stimulating innovation
and the use of more skilled workers.

Keywords: total factor productivity; innovation; skilled labour; firm size.
JEL: D24; J24; L25.

Acknowledgments: The authors acknowledge financial support from the
Ministerio de Ciencia y Tecnología, Programa Nacional de I+D+I, SEJ2005-
07814/ECON. Thanks are also due to the Fundación Empresa Pública for
supplying the data.






____________________
♠ Corresponding author. Email address: lcastany@ub.edu; Tel: +34 915450929; Fax: +34 934021821.
♣ Email address: elopez@ub.edu; Tel: +34 934037041; Fax: +34 934021821.
♦ Email address: rmoreno@ub.edu; Tel: +34 934037042; Fax: +34 934021821.

Page 2

Research Institute of Applied Economics 2007 Working Papers 2007/16, 24 pages
1
1. Introduction

Firm size has been considered a major source of heterogeneity, implying that small and
medium-sized enterprises (SMEs) and large firms will show a marked disparity in their
strategic decisions or in their productivity levels. Indeed, the evidence is compelling that large
firms are more productive than SMEs. According to the literature, the main reason for this
finding is a scale economies effect. SMEs encounter serious difficulties in achieving
economic results that are as good as those recorded by large firms and in accessing the main
factors that contribute to a firm’s productivity. See for example, Bartelsman and Doms (2000)
or Ruano (2002), who conclude that smaller firms tend to be less productive. However, SMEs
are seen as new firms starting out on their economic activity, which involves a high degree of
risk. Seen in this light, those that perform well survive and grow, while many others tend to
disappear. The probability of survival is not high, but those that survive are highly productive
and competitive, and represent an important source of economic growth and employment
(Audretsch, 2002).
Arguments supporting the positive role played by innovation and human capital in a
firm’s productivity can be found in, for instance, Griliches (1979), Crépon et al. (1998),
Griliches and Regev (1995), Haltiwanger et al. (1999) and Huergo and Jaumandreu (2004a).
However, small firms are usually considered to innovate less than large firms and to employ
fewer qualified employees (Schumpeter, 1942; Evans and Leighton, 1989; Acs et al., 1994;
Zábojník and Bernhardt, 2001; Huergo and Jaumandreu, 2004b). The difficulties small firms
face in accessing similar levels of innovation and human capital to those enjoyed by large
firms may limit their ability to achieve higher productivity levels. In addition, it can be argued
that the level of productivity is not only related to the efforts expended in technological
activities and human capital (among other characteristics of the firm), but also to the returns
that firms obtain from such efforts, in other words, to the impact of the use of innovation and
human capital on a firm’s productivity. Geroski (1998) argues that a firm’s size may have an
indirect effect on its productivity by conditioning the effect of other variables on productivity.
That is, SMEs and large firms might present different patterns of behaviour in relation to
innovation and human capital, the two variables of interest to us here. This author suggests
controlling for this indirect effect by analysing the coefficients of small and large firms
separately in the regression. Accordingly, differences in productivity levels between SMEs
and large firms would be observed if returns were lower in the former, regardless of the
intensity of use.

Page 3

Research Institute of Applied Economics 2007 Working Papers 2007/16, 24 pages
2
Building on these arguments, the hypothesis presented in this paper is that the
productivity differences between SMEs and large firms are associated with two of the main
determinants of a firm’s performance: the human and technological capital that firms
incorporate. In addition, we suggest that the contribution of these factors in explaining the
productivity-size gap might not only be due to the fact that large firms make a more extensive
use of them, but also because large firms obtain higher returns from their investment in
human and technological capital. In other words, our assumption is that every innovation and
every additional skilled worker incorporated in an SME provides returns which are lower than
those obtained in a large firm. Thus, the lower return on these factors might also explain why
small firms are less productive and why they have less incentive to use them.
The main purpose of this paper is therefore to provide evidence that supports this
hypothesis for a representative sample of Spanish manufacturing firms. In this regard, it is
worth mentioning that the situation presented by the Spanish economy is of interest when
analysing the reasons behind productivity differentials between small and large firms,
particularly because the economy is characterized by a smaller average firm size and a lower
proportion of large firms than other advanced economies. According to the Observatory of
European SMEs,1 in 2000, only 0.1% of Spanish firms employed more than 250 workers. By
contrast, the EU-15 average is 0.2% and some of the most advanced economies in Europe,
such as Denmark, Finland, Sweden and the Netherlands, reach values between 0.4% and
0.5%. Although there are few firms with more than 250 employees, they account for 20% of
Spanish employment. The EU-15 percentage, however, stands at 34% while, in the
aforementioned advanced economies, between 40 and 50% of the workforce are employed in
these large firms. These data reflect the importance of SMEs in Spain compared to other
advanced economies.
As discussed above, small and large firms seem to obtain different economic results
and take different strategic decisions. From this perspective, a large share of small firms in the
economy could be associated with lower aggregate productivity, as well as lower innovative
effort and investment in skilled workers. In this context, firm size, innovation and human
capital may interact in accounting for the weak productivity performance in Spain. In fact, our
hypothesis concerning the role of differences in returns implies that firm size conditions the
effect of innovations and employees’ qualification on productivity, so that size indirectly
affects productivity. The confirmation of this hypothesis has important implications for the
effectiveness of policies designed to improve productivity levels in SMEs through the
stimulation of innovation and the use of more skilled workers. In line with our assumption,

Page 4

Research Institute of Applied Economics 2007 Working Papers 2007/16, 24 pages
3
such policies would be more effective if they could simultaneously increase the returns that
SMEs are able to obtain from the use of further quantities of these two factors. A firm in its
initial stages, which is usually the case of SMEs, requires considerable levels of investment.
This finance is often very difficult to obtain given the high risks associated with a preliminary
project. Theoretically, the higher the risk is, the higher the returns from the investment will
be. But in practice, this mechanism does not often work and high risk projects generate lower
returns.2 The results for our sample of Spanish firms reveal that many of the differences in
productivity across firms of different size originate from differences in returns to innovations
and human capital, thus, lending support to our hypothesis.
The rest of this paper is outlined as follows. After this introduction, in which we have
presented theoretical reasons to show how small and large firms adopt different patterns of
behaviour in relation to productivity, innovation and human capital, we present, in section 2,
the total factor productivity (TFP) measure computed in this paper and describe the database.
Section 3 offers a descriptive analysis used in checking for differences in the TFP levels
between SMEs and large firms conditioned by the use they make of innovation and human
capital. In section 4 we further this analysis by estimating the impact of the knowledge
variables in order to ascertain whether the returns that the subsamples of small and large firms
obtain from innovation and human capital are different. The last section concludes with the
paper’s main findings and discusses some policy implications.

2. TFP Measure, Dataset and Variables
We use a sample of Spanish manufacturing firms drawn from the survey Encuesta sobre
Estrategias Empresariales (ESEE). This survey is an unbalanced panel that covers the period
1990-2002 and reports information on strategic decisions and the behaviour of firms. Firms
answered a comprehensive questionnaire every four years and a reduced questionnaire in the
intervening years (covering those issues that allegedly change annually), so that complete
information is available for 1990, 1994, 1998 and 2002. The reference population of the
ESEE is a sample of firms with 10 or more employees, whose activity corresponds to
divisions 15 to 37 of NACE-93, excluding division 23 (activities related to oil refinement and
fuel treatment). During the initial period, all firms with more than 200 employees were
required to participate (70% did). Firms with 10 to 200 employees were sampled randomly
according to industry and four size strata, retaining about 5%, in order to guarantee
representativity for every industry and firm size. The ESEE is designed to change as the
composition of the industry evolves. Newly created firms are selected using the original

Page 5

Research Institute of Applied Economics 2007 Working Papers 2007/16, 24 pages
4
criteria. Due to death and attrition, some firms were replaced by others in their industry and
size group so as to maintain representativity.3
Between 1990 and 2002, the ESEE has 37,141 observations, for 3,462 different firms.
However a number of these observations were lost in our analysis. First, because some firms
do not respond to some of the fields in the questionnaire that are necessary for our analysis;
and second, because we removed anomalous observations following the same criteria as those
adopted in the study by Ornaghi (2006) using the ESEE. Eventually, we obtained a sample of
13,035 observations over 13 years (1990-2002), for 2,104 different firms. Given that we have
an unbalanced panel, this means that we have around 800-1000 observations per year.
In our analysis we used TFP measured according to the index developed by Good et
al. (1996), which is derived from a translog production function.4 Its analytical expression for
a firm f in period t is as follows:

∑∑
=
s
2
∑
=
s
∑
=
i
=
−−−
−+−−+
−+−−=
tn
i
si issi is
t
ss
n
it iftit iftt ft ft
XXSSYY
XXSSYY TFP
1
1,1,
2
1
1
) ln ln)((
2
1
) lnln(
) ln)(ln(
2
1
) ln (lnln

(1)

where Y and Xi are respectively quantities of output and inputs i=1…n, Si is the cost-based
share of input i and the bar over the variables refers to their average. The first two terms on
the right hand side of the equation are the deviation of the firm’s output and inputs from those
of a hypothetical firm, which is the reference point in year t. The last two terms are the
cumulative change in the output and input reference points between year t and the initial year.
This second part introduces a productivity differential for each year (as output, inputs and
shares may change) and, therefore, accounts for possible technological changes. The
productivity index for a given firm and year is expressed in relation to the hypothetical firm in
the base period. Following Hall’s (1990) suggestion, weights are calculated as the share of
every input in the total cost of inputs. Appendix 1 provides a description of the measurement
of the variables needed to construct this TFP index.
As discussed in the introduction, we seek to investigate the role played by innovative
activity and human capital in the productivity levels of small and large firms. Many databases
and studies consider SMEs as those firms with fewer than 250 employees. However, our
database makes the distinction at 200 employees and uses different sampling schemes for the
two groups. Therefore, we will consider SMEs as those with 200 or fewer employees. We

Page 6

Research Institute of Applied Economics 2007 Working Papers 2007/16, 24 pages
5
consider it more appropriate to use the same criterion so as to guarantee representativity by
size strata.5
In line with previous studies of Spanish manufacturing firms (Huergo and Jaumandreu
2004a), we use process innovation as the measure of a firm’s innovative activity, on the
assumption that it is process and not product innovation that improves the mechanisms
through which inputs are transformed into output (Ornaghi, 2006).6 Specifically, a firm’s
innovative activity is defined as a dichotomous variable (Innovation) that takes value 1 if the
firm has implemented a process innovation. Process innovations are assumed to take place
when the firm gives a positive response to the following request: “Indicate if your firm
introduced some significant modification in the production process (process innovation). If
the answer is yes, please indicate the means: (a) introduction of new machines; (b)
introduction of new methods of organization; (c) both”.
Human capital (Skilled workers) is measured in terms of the formal education of the
labour force. This variable is defined as the ratio of skilled workers according to educational
level. Specifically, it includes engineers, graduates, middle level engineers, experts and
qualified assistants.

3. Differences in the use of Innovation and Human Capital and their effect on TFP
Using the variables described in the previous section and the index in expression (1), we
calculate TFP for a sample of Spanish manufacturing firms over the period 1990-2002. We
will first show that there are significant differences in the TFP levels between SMEs and large
firms. Then we will seek to verify whether these differences are conditioned by the use firms
make of innovation and human capital.

3.1 TFP levels by Firm Size
With regards to average TFP, the figures in Table 1 clearly confirm that productivity in large
firms is higher than that in their smaller counterparts, with differences being statistically
significant in each year (the t-test of equality of means in the last column of the table rejects
the null hypothesis that small and large firms have the same average TFP). However,
differences in TFP between small and large firms tend to become less pronounced over time
and the gap is narrower at the end of the period under analysis. This reflects the higher pace
recorded by the productivity growth of small firms since the mid nineties. The general
evolution in TFP for small and large firms shows an increase over time although there is a
deceleration during the second half of the nineties. In contrast with the first half of the

Page 7

Research Institute of Applied Economics 2007 Working Papers 2007/16, 24 pages
6
nineties, in which growth rates in small and large firms were quite similar (yearly average of
2.44% and 2.66% respectively), since the mid nineties small firms became more dynamic (an
annual TFP growth rate of 0.9% versus 0.4% in large firms). Thus the deceleration in
productivity growth was much more marked in the case of large firms.

[Insert Table 1 about here]

It can also be observed how the dispersion in the distribution of TFP, as measured by
the standard deviation, is higher for SMEs than it is for large firms, and that it increases over
time, corroborating the belief that SMEs constitute a highly heterogeneous group.
Interestingly, the figures reveal that there are SMEs with TFP levels that are even higher than
those in the most productive large firms. This can be explained by the existence of high-
growth firms or gazelles, which are in the spotlight of industrial policies. And this feature
seems to be increasing over time. In sharp contrast, the TFP levels for the less productive
SMEs are well below those of the less productive large firms, indicating that SMEs face
major difficulties and that some of them might end up exiting the market. And here as well,
the gap seems to be widening over time. In conclusion, it seems clear that on average large
firms are more productive. However, it should be borne in mind that i) dispersion in the TFP
distribution for both firm types increases over time, which can be read as an indication of the
boosting of the firms’ heterogeneity as regards their level of TFP, and ii) the less productive
firms in the Spanish manufacturing industry are SMEs, yet some of these firms in fact
perform better than the most productive large firms.

3.2. Innovation, Human Capital and TFP by Firm Size.
The share of firms in the total sample and in the two groups that introduced at least one
process innovation in each of the years under analysis is reported in Table 2. It can be
observed that around one third of the firms in our sample introduced new processes and that
this proportion did not seem to increase over time. Differences in innovative activity by firm
size can also be clearly deduced: around half the large firms obtained process innovations,
compared to just a quarter of small firms and these differences are statistically significant (as
shown by the test of equality of proportions). This result is consistent with the general finding
that large firms are more innovative.7 Table 2 also reports the share of firms that employ a
proportion of skilled workers above the median.8 In this case too, the figures reveal that SMEs
make a significantly less intensive use of skilled labour (as confirmed by the significance of

Page 8

Research Institute of Applied Economics 2007 Working Papers 2007/16, 24 pages
7
the test of equality of means), and that the difference in favour of large firms remains stable in
the period under analysis. The average percentage of qualified workers for the total sample is
around 8% at the beginning of the period and increased over time reaching 10% by the end.
For small firms, the percentage increased from 7 to 9%, and for large firms from 10 to 12%.
This result is in line with the general finding that large firms employ more qualified
employees.
[Insert Table 2 about here]

The figures in Table 2 thus confirm that small firms made a less intensive use of
innovation and human capital. If these two factors affected the level of productivity, we
would observe higher TFP in firms that made a more intensive use of the two factors, and
smaller differences in TFP between SMEs and large firms for firms using the factors with
similar intensity. The first panel in Tables 3 and 4 provides evidence concerning this first
issue for innovation and human capital respectively. The other two panels in these tables
verify whether there are significant differences between SMEs and large firms when the
intensity in the use of factors is controlled for.
In the case of innovation, in accordance with the theoretical arguments and previous
empirical evidence, Table 3 shows that firms which obtained process innovations were more
productive. In fact, the t-tests of equality of means strongly reject the null that innovating and
non-innovating firms had equal TFP levels, indicating that innovative firms were significantly
more productive. Thus, the lower innovative propensity in SMEs explains, in part, the lower
productivity levels for these firms. However, when comparing innovative SMEs and large
firms, some differences in TFP remain after controlling for innovative propensity. But the gap
was narrower than that observed for non-innovative SMEs and large firms, and decreased
over the period in such a way that in 2002 it was only significant at 10%. The evidence thus
suggests that differences in TFP associated with size are more important in the group of non-
innovative firms than they are in that of innovative firms. Seen in this light, innovation seems
to mitigate the differences in TFP between small and large firms.

[Insert Table 3 about here]

Similarly, it is interesting to investigate the effect of process innovations on TFP when
controlling by firm size. In order to do so, we compared the productivity of SMEs that
obtained new process innovations and those that did not. The second and third panels in Table

Page 9

Research Institute of Applied Economics 2007 Working Papers 2007/16, 24 pages
8
3 show that the average TFP of small innovative firms is much higher than that of small, non-
innovative firms, and that the differences are significant throughout the period. Thus, small
firms obtaining a process innovation were able to achieve higher TFP levels than those SMEs
that did not innovate. In sharp contrast, the TFP gains derived from obtaining a process
innovation in the case of large firms was almost negligible. In fact, the TFP gap between large
innovative and non-innovative firms is not significant, except in 1998 at 5%. Since the gains
in productivity associated with process innovation are more important in small than they are
in large firms, these results suggest that obtaining process innovations may be a key element
in helping small firms increase productivity and become more competitive.
The results of an analogue analysis for the use of human capital are summarised in
Table 4. It can be observed from the first panel in the table that the differences in TFP
between firms that have a high proportion of skilled workers (above the median) and those
employing a low proportion (below the median) are quite important. As expected, firms that
employ more skilled labour are significantly more productive (as confirmed by the t-tests of
equality of means). However, the TFP gap between firms of different size does not vanish
when considering firms that make an intensive use of skilled workers. The second panel of
Table 4 shows that among firms that employ a high proportion of qualified workers, large
firms are significantly more productive than their smaller counterparts. However, this
statement should be qualified: the differences in TFP seem to decrease over time and they are
appreciably smaller than those observed for the group of firms with a low proportion of
qualified workers (third panel of Table 4). Finally, a comparison of TFP figures in the second
and third panels in Table 4 confirms that the employment of skilled labour plays a role in
explaining TFP differentials within SMEs and, in contrast with the case of innovation, also
within large firms.
[Insert Table 4 about here]

From the descriptive analysis conducted to this point we can conclude that differences
in the use of innovation and human capital between SMEs and large firms alone cannot fully
account for the productivity-size gap. The fact that after controlling for the amount of
innovation and human capital, SMEs are still significantly less productive than large firms
supports our hypothesis that they might be obtaining lower returns from the use of these
factors. But this conclusion is not definitive as differences related to size within groups of
firms with similar innovative activity and similar levels of employment of skilled labour
might well be caused by other well-known determinants of a firm’s productivity.

Page 10

Research Institute of Applied Economics 2007 Working Papers 2007/16, 24 pages
9

4. Differences in Returns to Innovation and Human capital
In this section we further the analysis by investigating whether the returns to the use of
innovative activity and human capital differ between SMEs and large firms. In so doing we
simultaneously account for the effect of these two factors, and for a number of others that
have been shown to affect a firm’s productivity. Our empirical specification is quite similar to
that adopted in Griliches and Regev (1995), where they estimate a production function at the
firm level by including measures of human and technological capital. Instead of the
production function, we use the TFP index described in (1) as our dependent variable and
innovation and skilled labour as the explanatory variables, whose effects on productivity we
wish to evaluate. Hence, the empirical model can be expressed as follows:
ftft1 ft21 ft10 ft
uZ HK INN TFPln
++++=
−−
γβββ
(2)
where lnTFPft is the logarithm of the total factor productivity index in firm f in year t, INNft-1
is a dummy variable that takes value one if firm f reports to have obtained a process
innovation in year t-1, HKft-1 is the proportion of skilled labour for firm f in year t-1, and Zft is
a set of standard control variables: firm size9, age, industry and year effects, and u is an error
term. Firm size (Size) is defined as the log of the total number of employees. The age (Age) is
defined as the number of years since the firm was set up, whereas the sectoral heterogeneity is
controlled for by a set of 20 dummy variables (Sector dummies) according to the NACE-93
classification, where the omitted category is “Other manufacturing industries”. Finally, a set
of year dummies is included to control for exogenous technical progress and effects of the
business cycle that are common to all firms (Year dummies).
The possible endogeneity problems in labour, capital and materials that appear in the
production function estimations are avoided when calculating a TFP index and using input
prices instead of estimating their returns to calculate the participation of each input in the
production function. Still, the exogeneity of innovative activity and human capital in a
specification such as that in (2) can be questioned. In the absence of appropriate available
instruments, we have used the lag of the variables instead of their contemporaneous value to
mitigate the effect of endogeneity.10 Specifically, for innovation and human capital we have
considered the values in the previous year.11
In addition to the baseline specification described in (2), we have also estimated the
returns to our variables of interest from a specification that includes additional control
variables. The idea is to capture other sources of observed heterogeneity in a firm’s

Page 11

Research Institute of Applied Economics 2007 Working Papers 2007/16, 24 pages
10
performance. The variables included in Z for the robustness analysis are basically controlling
for the ownership structure, for the degree of competition faced by the firm and its market
orientation, the productive capacity used by the firm, for the region in which it is located and
for the economic cycle. As variables related to the ownership structure of the firm, we
introduce the proportion of foreign-owned capital of the firm (Foreign capital), the proportion
of publicly-owned capital of the firm (Public capital) and a dummy as to whether the firm
belongs to a group of firms (Group). To approximate the competition faced by the firm, we
include a set of dummy variables on the geographical scope of the firm’s main market
(Market dummies). This considers whether the market is local, provincial, regional, national,
international and a category that includes all the previous categories, which is the omitted
category. Exports are measured as the log of the value of exports expressed in 1990 constant
pesetas (Exports). The productive capacity of the firm is a question directly put to firms in the
survey (Productive capacity). The region of the firm is a set of 17 dummy variables for the
NUTS II regions (Region dummies). The omitted category is “La Rioja”. Finally, it should be
mentioned that all the estimates include random effects to account for unobservable
heterogeneity among firms.12
Table 5 summarises the results of the estimation of the aforementioned specifications
for the total sample of firms and for the group of SMEs and large firms separately. In all
cases, the Lagrange Multiplier test rejects its null hypothesis of absence of unobservable firm
heterogeneity, confirming the appropriateness of the random effects estimation over a
specification excluding those effects. Controls for sector, region and year are clearly
significant as well. Results obtained by using the total sample of firms —column (i)—
confirm that the effect of the two variables of interest, innovative activity and skilled workers
is positive and significant. This confirms that in our sample of firms the knowledge capital
acquired by a firm improved the mechanism by which inputs are transformed into output.
Process innovations reduced the unitary cost of production, and then caused productivity
increases. However, the effect seems to be modest: shifting from being a non-innovative to
innovative firm increased the TFP by 2%.13 The positive and significant coefficient for human
capital proves that a more intensive use of skilled labour increases productivity because
workers can do any task that requires something more than just the simple workforce in a
more efficient manner. In fact, a one-point increase in the ratio of skilled workers increases
TFP of the average Spanish manufacturing firm by 15%.14 The estimate of the effect of both
variables is quite robust to the inclusion of additional controls for a firm’s heterogeneity, as
deduced from the results in column (iv). The only change worth mentioning is the decrease in

Page 12

Research Institute of Applied Economics 2007 Working Papers 2007/16, 24 pages
11
the estimate of the returns to human capital. However, it should be borne in mind that most of
the control variables ought to capture the channels by which human capital contributes to
increase the productivity, thus causing a reduction in the estimate of its effect.

[Insert Table 5 about here]

As for the difference in the estimated effect of innovation and human capital between
SMEs and large firms, results in columns (ii) and (iii) clearly indicate that the strength of their
contribution to productivity enhancement varies in each group of firms. Whereas the
coefficient of innovation in large firms is much higher than that estimated for the total sample,
the effect of this variable for the SMEs seems to be negligible. Large innovative firms have
TFP levels that are almost 4% higher than those of the large firms that do not report having
adopted any innovations. In sharp contrast, our results reveal that once we control for other
variables that affect a firm’s productivity, the TFP gap between innovative and non-
innovative SMEs is not significantly different from zero.
Differences in the estimated effect of human capital between SMEs and large firms are
remarkable as well, although in this case the return to skilled labour remains significant in
SMEs. A one-point increase in the ratio of skilled workers in an SME increases its TFP by
12%. But this increase rises to 20% if it is a large firm that increases its ratio of skilled labour.
The results from a formal test (not reported here for reasons of space) confirm these
arguments.15 The test rejects the null hypothesis of homogeneity in the effect of innovation
and human capital in the SMEs and in the large firms with a p-value of 9%. Individual tests
for the significance of the effects of each of the variables in isolation reveal that the evidence
against similar returns in firms of different size is stronger in the case of innovation
(significant at 5%) than in that of human capital (significant at 10%).
The inclusion of additional controls in columns (v) and (vi) only modifies the
estimated effect of human capital in SMEs and in large firms. In small firms, the estimate of
the effect falls dramatically (becoming not significant). A decrease is also recorded for large
firms, although of a much lower magnitude. In any case, the difference between the point
estimate of the parameter of skilled labour in SMEs and in large firms is even larger than in
the baseline specification. Therefore, we can conclude that the difference in the estimated
returns to innovation and human capital between firms of different size should not be assigned
to the omission of other observable variables affecting a firm’s productivity.

Page 13

Research Institute of Applied Economics 2007 Working Papers 2007/16, 24 pages
12
As in the baseline specification, a formal test rejects the null hypothesis that returns to
innovation and human capital are the same in SMEs and in large firms. And in this case the
evidence seems to be stronger as the p-value of the Wald test statistic is now 5%.
To summarize, both innovation and human capital seem to play a role in enhancing a
firm’s productivity, though the evidence suggests that the magnitude of these effects is very
closely related to firm size. In fact, after controlling for a large set of conditioning variables
and accounting for firm heterogeneity, the empirical evidence in this section suggests that the
effect of innovation and human capital on productivity is only marginal in the case of the
SMEs, while it is far from negligible for large firms. Thus, it seems that SMEs do not only
make a less intensive use of these knowledge factors, but they also obtain much lower returns
from them. As a consequence, it can be concluded that some of the TFP gap between SMEs
and large firms might well be caused by the difference in their returns to innovation and
human capital.

5. Conclusions
Starting with the generally accepted belief that innovation and human capital play a crucial
role in improving a firm’s performance, this paper analyses whether the two factors have a
different impact on SMEs and large firms, and might therefore be identified as a possible
explanation for differences in TFP levels between these two firm types.
The descriptive analysis conducted here supports the hypothesis that the TFP
differences between small and large Spanish manufacturing firms are due not only to
differences in the level of use of knowledge capital, but also to differences in the effect that
this capital has on TFP. The fact that after controlling for innovation and human capital,
SMEs are still significantly less productive than large firms seems to suggest that the former
might be obtaining lower returns from the use of these factors. But we cannot draw a definite
conclusion as differences related to size among groups of firms with similar innovative
activities and similar levels of employment of skilled labour might well be caused by other
well-known determinants of a firm’s productivity, which should be analysed by conducting a
regression analysis.
After controlling for a large set of conditioning variables and accounting for firm
heterogeneity, both innovation and human capital seem to play a role in enhancing a firm’s
productivity. However, small and large firms follow different patterns of behaviour in relation
to innovation and human capital: large firms obtain positive and significant returns on their
investments in relation to these factors, which are significantly higher than they are for small

Page 14

Research Institute of Applied Economics 2007 Working Papers 2007/16, 24 pages
13
firms. As a consequence, it can be concluded that some of the TFP gap between SMEs and
large firms might well be caused by the difference in their returns on innovation and human
capital.
The effect of innovations on small firms was found to be only marginal and not
statistically significant when including additional control variables. The higher returns on
innovation in the case of large firms might be explained by the easier appropriability of
returns on innovation in the case of these firms. According to Klepper (1996) and Cohen and
Klepper (1996), the larger the firm, the more output over which process R&D fixed costs can
be averaged, implying that returns to process innovations are higher, which encourages
additional innovative effort. In this view, economic policies focused on increasing the
innovative activity for small firms are important, as we have observed how the productivity
gap between small and large firms becomes narrower for innovative firms. However, it would
only have a relevant impact if SMEs improved their returns on innovation; otherwise an
additional innovation in an SME would have a smaller impact on TFP than in a large firm and
a certain gap would remain.
On the other hand, the returns derived from employing qualified workers are larger in
the case of large than small firms. These higher returns to human capital in large firms can be
explained by the fact that the costs of communication related to the absorption of new
information can be somehow attenuated by specialization, and large firms are more likely to
specialize (Bolton and Dewatripont, 1994). Thus, economic policies focused on stimulating
the more intense use of qualified labour force in small firms would only make sense if the
returns on human capital in these firms could be improved, that is, if they could take more
advantage of their investment in human capital.
Finally, these results can be added to the previous literature that has analysed the role
of technological and human capital in improving productivity, with our additional emphasis
on the role of returns derived from investments in these two factors. In agreement with the
literature reporting on the technological gap between Spain and its European neighbours, we
find that innovation may play an important role and increase technological levels, leading to
productivity improvements for the industry as a whole. Our results are also in line with the
recommendation of the National Reform Program in the Lisbon Agenda that Spain increase
its human capital levels. We find that increasing human capital in small firms can improve
their productivity, but not to the same degree as in large firms. Thus, obtaining additional
innovations and increasing the proportion of qualified workers in small firms would only have
a positive impact on productivity if the returns of these firms increased. If this is not the case,

Page 15

Research Institute of Applied Economics 2007 Working Papers 2007/16, 24 pages
14
this effort would have a limited impact on a small firm’s productivity and, thus, on the
industry as a whole.

Appendix. Variables Involved in the TFP Index
All the quantities used to compute the TFP index are expressed in thousands of constant
pesetas of 1990, except for labour, which is measured as the number of hours worked.
We follow the same criteria as that used by Martín-Marcos and Suárez-Gálvez (1997),
Suárez-Gálvez (2001), Martín-Pliego et al. (2001), Delgado et al. (2002), Aw et al. (2003),
Huergo and Jaumandreu (2004a) and Huergo and Moreno (2006).
The output is defined as the production of the firm (measured as sales plus the
variation of stocks for sale). To deflate the nominal production, we construct a firm-specific
price index. The ESEE offers information on the price increases in the five main markets
where firms operate. The price index for output is calculated on the basis of the weighted sum
of the price increases in the different markets where the firms operate, where weights are the
sales in each market.
Labour input. It is calculated as the total effective hours of work, which is obtained by
multiplying the total number of employees (full time employees plus part time employees
divided by two plus the number of temporary employees) by the effective hours worked
during the year (normal hours plus overtime minus hours paid but not worked).
Capital Input. To obtain an estimation of the stock of capital at replacement cost we
use the permanent inventory method, which consists of calculating the stock of capital for an
initial year and, for the subsequent years, subtracting the depreciation, adjusting the prices to
take inflation into account, and finally, adding the flows of gross fixed capital formation that
have taken place over the year under consideration. The stock of capital in the initial year is
calculated on the basis of the balance sheets. More detail on methodology can be found in
Martín-Marcos and Suárez-Gálvez (1997).
Intermediate inputs. The amount of intermediate inputs in nominal terms includes the
purchases (acquisition of raw materials purchases, energy, etc.) and external services minus
the variation in stocks of purchases. To deflate these nominal quantities, we use a firm-
specific price index, which is calculated on the basis of the price increases of these inputs
(weighted by the share of their cost on the total cost).
To calculate the shares of inputs we use the percentage of their cost on the total cost of
inputs. The cost of labour input is calculated as the personnel costs of firms, which is deflated
using the consumer price index. The cost of capital is measured as the user cost of capital, that