• Non ci sono risultati.

Financial Constraints to Innovation in the UK: evidence from CIS2 and CIS3

N/A
N/A
Protected

Academic year: 2021

Condividi "Financial Constraints to Innovation in the UK: evidence from CIS2 and CIS3"

Copied!
20
0
0

Testo completo

(1)

Financial constraints to innovation in

the UK: evidence from CIS2 and CIS3

By Alessandra Canepa* and Paul Stonemany

*Economics and Finance, School of Social Sciences, Brunel University yWarwick Business School, University of Warwick, Coventry CV4 7AL; e-mail: paul.stoneman@warwick.ac.uk

The role of financial factors as constraints to innovation in the UK is explored using data on individual returns to the second and third Community Innovation Surveys. It is found that financial factors do impact upon innovative activity and that impact is more severe in higher tech sectors and for smaller enterprises. These results extend but largely confirm the results in the extant literature by using a different approach to the standard cash flow based method, encompassing a wider class of investment phenom-ena (innovation rather than just R&D) and exploiting a new data base.

JEL classification: O3.

1. Introduction

Recent literature, for example, Carpenter and Petersen (2002a), Basu and Guariglia (2002), and Bond et al. (2003), has addressed the proposition that many dimen-sions of firm behaviour and performance may be affected by financial and/or liquidity constraints. A particular strand has considered the impact of such constraints upon the investment behaviour of firms (e.g. Hubbard, 1998), including a more limited literature, surveyed by Hall (2002), looking at the impact of finan-cial constraints upon R&D expenditure. The majority of this literature argues that a changed availability of internal funds would not much affect investment (or R&D) if external funds were not constrained and thus the firm is only likely to be financially constrained if the firm’s investment in plant and machinery or R&D is particularly sensitive to cash flow. Tests for constraints thus proceed by including cash flow covariates in a standard investment (or R&D) equation. However the validity of this approach is much disputed. Papers by Kaplan and Zingales (1997), Bond and Cummins (2001), Bond et al. (2004), and Cummins et al. (2006), argue the point with the current state of the argument reflecting the view of the latter papers that cash flow sensitivities are unable to reflect financial constraints in an unbiased manner, largely on the grounds that, as an indicator, ß Oxford University Press 2007

All rights reserved

Oxford Economic Papers 60 (2008), 711–730 711 doi:10.1093/oep/gpm044

(2)

cash flow may reflect investment opportunities that are not properly accounted for in the modelling.

This paper is concerned with the potential impact of financial constraints on firms’ investments in innovative activity. The paper is itself innovative in that it explores a data set that as far as we are aware has not been used to this end before. This data encompasses individual firm responses to questions in two UK1 Community Innovation Surveys (CIS) relating to the existence of financial con-straints and their impact upon firm innovative activity. This data has two further distinct advantages. First, its use enables one to avoid the controversies surrounding the use of cash flow methodologies. Second, in the CIS surveys innovative activity is defined to include the costs of the acquisition of new capital goods, licensing fees etc. as well as R&D per se and given that the more commonly analysed R&D is only one aspect of innovative behaviour (reflecting generation of technology rather than use) this wider concept is to be preferred.

Stiglitz and Weiss (1981) consider a firm to be credit rationed if it does not get as much credit as it wants although it is willing to meet the conditions set by the lender on equivalent credit contracts. According to Hall (2002) a financial constraint is said to exist when, even if there are no externalities involved in the firm’s investment activity, there is a wedge (perhaps even a large wedge) between the rate of return required by an entrepreneur investing his own funds and that required by external investors. In essence therefore a firm is considered credit or financially constrained if it cannot raise external funding at the market price or in order to raise external funding it has to pay over the market price.

It is unnecessary to repeat the full theoretical grounding for propositions relating to the existence of financial constraints. This can be found elsewhere including Bond et al. (2003) and Hubbard (1998). The main foundations of the theory are based on asymmetric information between firms and the suppliers of finance with associated moral hazard and adverse selection problems. In particular small firms may be more prone to financial constraints as a result of indivisibilities in search costs, shorter track records and less collateral (Bond et al. 2003). High tech firms (Carpenter and Petersen, 2002b) by their nature may offer riskier investments (with innovation of a type less likely to have been undertaken elsewhere and thus with particular problems in observing systematic risk), greater information asymmetry, shorter track records, less collateral and assets that are less realisable.

Hall (2002) presents a survey of empirical evidence relating to the theoretical propositions on the impact of financial constraints on R&D. She considers that the existing literature based on the cash flow methodology (whether valid or not) yields conclusions that largely support the theoretical propositions, including that small firms are more likely to be financially constrained while the evidence for a financing gap for large and established firms is less compulsive, and that firms ...

1We have also undertaken some inter country analysis the results from which are available from the

authors upon request.

(3)

(especially small and start up) in R&D intensive industries face a higher cost of capital. Following in these footsteps, in this paper an analysis is undertaken of the responses to questions in the CIS asking whether the firm considered its innovative activity to be financially constrained, according to size and industrial sector.

The next section provides detail upon the CIS itself, the questions included within the CIS relevant to financial constraints and the data upon responses that is available for study. Section 3 presents data on the overall patterns of responses and discusses further how the responses to the CIS questions may be interpreted. Section 4 provides details of an econometric analysis of the data. Conclusions are drawn in Section 5.

2. The Community Innovation Survey: data description

Details on the UK applications of the CIS surveys are provided in the Appendix. We make use of individual firm response data (for which we are very grateful to the Department of Trade and Industry for providing us privileged access) from two surveys which we label CIS2 and CIS3. Both the CIS2 and the CIS3 are cross-section data sets, the former relating to the period 1994–96 the latter to the period 1998–2000. We concentrate upon whether the patterns of constraints revealed by the survey data do in fact reflect that enterprises2that are small or in higher tech sectors are more constrained in their innovative activity by financial factors than other firms, as theory and prior empirical evidence suggest.

The section of the CIS questionnaire of most interest in this work concerns the factors hampering innovation. Although the CIS2 and CIS3 surveys were meant to address the same issues the actual questions asked in the two surveys were different in nature. In Fig. 1 we present the key question asked of firms responding in the two surveys. From the responses we wish to draw inferences as to whether the behaviour of firms that intended to innovate was in fact constrained by financial factors.

In CIS3 each sample firm was asked to grade the importance of particular constraints to its ability to innovate in terms of whether there was any effect or not. All firms were asked this question whether they actually innovated or not. It is clear that a number of the firms in the sample may have had no intention to innovate (although they may have had the ability). The wording of the question however— ‘grade the importance of the following constraints’— leads us to believe that such firms would tend to indicate no or low effect. We thus interpret medium or high effect as implying that the firm was intending to innovate and was con-strained by one of the factors listed. Even if that might yield some bias in absolute terms we think that it will not compromise any comparisons of the relative importance of different constraints. In the analysis of the CIS3 data therefore ...

2The CIS surveys sampled enterprises but some of its questions refer to companies. For the purpose of

this paper enterprises, companies, and firms are considered interchangeable terms for the reporting units.

(4)

CIS2

The innovation activity of your company could be hampered by various factors which might prevent innovation projects or slow up or stop projects in progress

a) Has at least one innovation project in 1994–1996 been yes no - seriously delayed

- terminated after being started - not even started

b) If yes on atleast one question, tick the relevant factors in the respective columns

Hampering factors Seriously delayed Terminated Not even started after being

started 1. Excessive perceived economic risks

2. Lack of availability of finance 3. Cost of finance (*) 4. Organisational rigidities 5. Lack of qualified personnel 6. Lack of information on technology 7. Lack of information on markets 8. Direct cost of innovation (**) 9. Fulfilling regulations, standards 10. Lack of customer responsiveness

11. Lack of opportunities for co-operation with other firms or technical organisations 12. Lack of availability of external technical services

13. Other (Please specify)

(*) Cost of finance refers to the costs incurred in obtaining any finance that may be available for innovation activities. This may be a hampering factor if, for example, interest rates are too high, or repayment conditions unfavourable.

(**) Direct cost of innovation refers to the costs incurred in carrying out technological innovation. This may hamper innovation if, for example, machinery or materials necessary for innovation are too expensive.

Fig. 1. Obstacles to innovation: question asked in the CIS2 and CIS3.

(5)

our sample is the original entire sample population of 8172 enterprises, less 932 enterprises that had to be discarded due to missing data points, leaving 7240 firms (of whom 2329 actually undertook innovative activity between 1998 and 2000).3

The responses to CIS2 are more difficult to interpret. The question is in two parts, (a) and (b), with only those firms that have suffered delay, termination or ‘not starting’ going on to answer part (b) re the hampering factors. Once again we wish to draw inferences as to whether the behaviour of firms that intended to innovate was in fact constrained by financial factors. The problem is again, however, whether all those enterprises that intended to innovate, and only those enterprises, answered the questions. It is possible to answer no to all of part (a) because the enterprise has no intention to innovate, and thus one cannot use the proportions of yes and no responses as a clear indicator of whether a firm has been hampered or not in its innovation activity. To overcome this problem, with CIS2 we proceed by considering only those firms that report (in another question) that they have undertaken innovative activity (product or process) in the survey period.

CIS3

A range of factors may inhibit your ability to innovate. Please grade the importance of the following constraints during the period 1998-2000: (Please tick one of the boxes in each row)

Importance

No effect Low Medium High Economic factors Excessive perceived economic risks

Direct innovation costs too high Cost of finance

Availability of finance

Internal factors Organisational rigidities within the enterprise Lack of qualified personnel

Lack of information on technology Lack of information on markets Other factors Impact of regulations or standards

Lack of customer responsiveness to new good or services

Fig. 1. Continued.

...

3Missing data mainly related to the factors hampering innovation (see Fig. 1). In order to test if data

were missing at random for each factor we created a dummy variable that took the value one if the data was missing and zero otherwise. Then, the Pearson’s 2test statistics were calculated in order to test for

the independency between each hampering factor and the missing group. In all cases the test statistics did not reject the null hypotheses.

(6)

We are well aware that there may well be firms that intended to innovate but, as a result of hampering factors, did not complete any projects, and so will not appear as innovative firms in our sample. To the extent that such firms exist our results on the proportion of firms whose innovative activity is affected by hampering factors will be biased downwards. We are also aware that this is likely to lead to the exclu-sion of firms pursuing only one innovative project, and thus bias the results against finding greater constraints to small firms. Thus to the extent that we find con-straints from financial factors and especially on small firms, these will probably be underestimates of the true effects. When we compare the relative importance of different hampering factors to each other in CIS2 the problems seem much less. The sub sample of firms in the CIS2 dataset that introduced new or significantly improved products onto the market between 1994 and 1996 contains 1016 firms (excluding the 1326 firms who did not innovate), and this is the sub sample on which we concentrate.

Exploring questionnaire responses is of course a different methodology to the standard cash flow approach, however both approaches face a similar problem, how to identify when intentions to innovate are constrained. The cash flow meth-odology uses an assumed investment model to predict innovation expenditures and then explores whether cash flow variables intermediate between actual and pre-dicted expenditure. The weakness is whether cash flow is an objective representative of the financial constraint. In the current methodology, the presence of a constraint or hampering factor is clearly indicated in the survey response, but the weakness is that it is much less clear which respondents had the intention to innovate and would have innovated in the absence of the constraint.

Perhaps the fairest stance is to consider the alternative methodological approaches as complementary with different costs and benefits. The CIS approach yields large samples and enables a wider definition of innovation than R&D. In addition it provides direct indicators of the number and type of firms that have experienced delay in, termination of, or failure to start an innovation project be-cause of a lack of available finance. Also, whereas the cash flow methodology only allows for constraints in the capital market, the CIS methodology allows for con-straints in other markets and of other kinds (for example, a firm may face a skilled labour constraint prior to a finance constraint and thus the financial constraint is not binding). However, the cash flow methodology, through control variables, can implicitly allow for the returns to investment activity, whereas the CIS approach cannot indicate whether projects for which finance is not available offer a market rate of return.

3. Questionnaire response: the overall picture and

further matters of interpretation

Tables 1 and 2 report the frequency rankings associated with each hampering factor in CIS2 and CIS3. The results are presented separately for manufacturing and services by firm size class and level of technology (high-tech, low tech).

(7)

Table 1 Ranking of factors hampering innovation for CIS2 dataset

Manufacturing

... ...Services Factors hampering innovation Small Medium Large Low-tech High-tech Small Medium Large Low-tech High-tech

Excessive perceived economic risks 4 2 1 7 7 3 1 1 1 –

Innovation costs too high 2 4 2 5 8 3 3 2 1 –

Cost of finance 3 3 4 5 5 2 4 4 3 –

Availability of finance 1 1 3 1 2 1 2 3 1 –

Organisational rigidities 7 6 6 3 8 4 7 6 3 –

Lack of qualified personnel 5 5 6 6 6 4 5 6 3 –

Lack of information on technology 7 9 8 2 3 4 6 8 3 –

Lack of information on markets 5 8 7 7 4 5 7 7 3 –

Fulfilling regulations, standards 7 9 8 4 9 5 7 8 3 –

Lack of customer responsiveness 6 7 5 1 1 4 7 5 2 –

Note: Sample size is 1016, with 823 firms in manufacturing, 193 in services and 304, 311, 401 in the small, medium, and large size bands respectively.

a.

canepa

and

p.

stoneman

717

(8)

Table 2 Ranking of factors hampering innovation for CIS3 dataset

Manufacturing

... ...Services Factors hampering innovation Small Medium Large Low-tech High-tech Small Medium Large Low-tech High-tech

Excessive perceived economic risks 2 2 2 10 10 2 2 2 10 10

Innovation costs too high 1 1 1 1 1 1 1 1 1 1

Cost of finance 3 5 4 2 2 3 6 7 2 2

Availability of finance 5 6 6 2 6 7 7 8 6 6

Organisational rigidities 6 4 8 9 9 10 8 6 8 9

Lack of qualified personnel 10 9 5 6 3 5 3 4 4 3

Lack of information on technology 9 10 10 8 8 9 9 9 7 8

Lack of information on markets 8 8 9 7 7 8 10 10 9 7

Fulfilling regulations, standards 4 7 7 5 4 4 5 5 3 4

Lack of customer responsiveness 7 3 3 3 5 6 4 3 5 5

Note: Sample size is 7240 with 3953 firms in manufacturing, 3287 in the service sector, and 4184, 1797, 1259 in the small, medium and large size bands respectively.

718

financial

constraints

to

(9)

The industrial classification is presented in Table A1 of the Appendix. There are 17 industry classifications (11 manufacturing sectors, six service sectors) and three firm size classes (small, medium, and large, defined respectively as firms with 20 to 49 employees in the manufacturing sector (10 to 49 in the service sector), 50 to 249 employees and more than 250 employees). The classification of industrial sectors according to the level of technology for manufacturing follows the OECD method (see Hatzichronoglou, 1997). The OECD classifies manufacturing indus-tries according to low, medium, medium-high, and high level of R&D intensity. We group the OECD low and medium-tech as low-tech sectors and medium-high and high tech as high-tech sectors. For the services sectors, to the best of our knowledge, no such classification exists. However, in the literature (see for example Young, 1996) services like telecommunication, financial intermediation, R&D, computers, and engineering are usually classified as high-tech whereas construc-tion, electricity, wholesale trade, and transport are regarded as more traditional sectors. This enables us to classify service firms in CIS 3 as high or low tech but the coverage of CIS 2 is such that we consider all service firms in CIS 2 as low tech.

For the CIS2 responses, an innovation is considered hampered if the project was ‘delayed’, ‘not started’, or ‘abandoned’. For CIS3 a factor has been considered to be hampering if the responding firm graded it as being of medium or high import-ance. Note that equal frequencies for different hampering factors correspond to equal ranking (for example, in Table 1, for small firms in the service sector the hampering factors are ranked up to five instead of 10).

The ranking of the obstacles to innovation by firm size in Table 1 shows that in CIS2 ‘costs of finance’ and ‘availability of finance’, which we temporarily jointly label financial factors, were important relative to other hampering factors. However there are differences across size bands. In both the manufacturing and service sectors, small enterprises rank the availability of finance as the most important factor hindering innovation (whereas for large enterprises it is excessive economic risk that matters most). For CIS3 the picture is more homogenous with ‘innovation cost too high’ ranking as the most important hampering factor across size bands and industrial sector, and excessive perceived risks coming in second. But costs of finance appear quite highly ranked even if differing in importance across firm size. However at this level of data aggregation the break down by level of technology does not show any particularly large or obvious differences between low-tech and high-tech sectors (note that for data reasons in CIS 2 the service sector firms are all considered low tech).

Although indicative, the findings that ‘costs of finance’ and ‘availability of finance’ are relatively important reported hampering factors is not the same as stating that financial factors (relatively) constrain innovative activity, or even more that they constrain innovative activity is the sense defined by Stiglitz and Weiss (1981) or Hall (2002). A statement that innovation activity has been postponed or delayed or cancelled because the cost of finance is too high may just be a different way of stating that the demand curve for investment funds is down-ward sloping, which is not ‘a constraint’ in the normal sense. An alternative way

(10)

of putting this is that a statement that the cost of finance is too high may just mean that at the (unconstrained) market rate some investment projects are not viable.

A statement that there is a ‘lack of availability of finance’ is probably a clearer indicator of there being financial constraints. Again however there is some problem of interpretation. The statement could mean (i) that at current market rates the firm cannot raise the necessary funding to proceed with innovation activity or (ii) at rates the firm is willing to pay innovation funding cannot be raised. The former might indicate financial constraints in the sense of Hall whereas the latter would not. We argue that if the latter situation is the case, then the firm is more likely to report that the cost of finance is too high. Thus only a response that innovation is constrained by ‘a lack of availability of finance’ is likely to indicate that the firm cannot raise the necessary funding at market rates. For the rest of this paper we thus consider that the firm faces a financial constraint if and only if it reports that there is ‘a lack of availability of finance’. It is of course still possible that a lack of availability of finance is the result of a project being non-viable and the financiers’ rejections of a project on these grounds being interpreted as a lack of available finance. This will be a problem with all survey data of this kind and cannot be overcome.

On this basis we now undertake some initial exploration of the validity of previous findings (see above) suggesting that smaller firms may experience more financial constraints. We also investigate further if high-tech sectors face more barriers to innovation than low-tech sectors. In Table 3 we report the relative frequencies of responses re the impact of the availability of finance by firm size and level of technology for CIS2 and CIS3. The Pearson’s 2 test for the hypothesis that the rows and columns in the two-way table are independent, and the likelihood ratio statistics (LR) are also reported, together with their respective p-values.

For CIS2 we tabulate whether the firm considered that the lack of availability of finance hampered innovative activity or not and whether the hampering was in terms of delay, termination or not starting. As seen in the last column, about 15% (100% minus 84.94%) of the sample (of 1016) considered that the availability of finance did affect their innovative activity (which may be downward biased, see above). There is thus evidence of financial constraints. An analysis of this raw data suggests that the lack of availability of finance mostly acts to delay firms’ innovation activity (rather than lead to termination or abandonment) across all size classes and technology levels. The observation that in larger firms projects are (relative to smaller firms) more likely to be terminated (rather than delayed or not started) as a result of a lack of available finance may reflect the fact that larger firms on average have higher levels of research and development spending and broader production programmes, and thus may have a greater like-lihood of engaging in risky projects; as a result they may be more likely to terminate projects.

By firm size class (304 small, 311 medium, and 401 large firms), the raw data indicates that the majority of those firms in the sample experiencing hampered

(11)

projects are in the small and medium sized class. However taking account of relative representation in the sample the probability that a firm reports that a project has been hampered by the availability of finance is 0.16% for small firms, 0.15% for medium sized firms and 0.14% for large firms which data hardly indi-cates large differences by firm size.

The break-down by level of technology shows that, in CIS2, firms in high-tech sectors are more likely to experience hampered projects. Indeed, out of 85% of the enterprises who did not report hampered projects about 50% were in the low-tech sector. Allowing for different representations in the sample 14% of low tech firms were hampered by the availability of finance and 18% of high tech firms. The major impact was again in terms of delays to the project.

For the CIS3 data (sample 7240, 4184 small, 1797 medium, and 1259 large firms), we report the grading by enterprises of the importance of the availability of finance dividing the responses in to ‘no or low’ and ‘medium or high’ effect categories. With about one third of the sample (32.69%) reporting that finance availability was of medium or high importance there is clear evidence that finance was a constraint on innovative activity.

About 20% of the firms declaring financial constraints as being of medium or high importance fall within the small size class. However, taking account Table 3 The impact of lack of availability of finance by firm size and level of technology (% of sample)

Small Medium Large Low-tech High-tech Total CIS2 No hampered projects 25.00 26.18 33.76 49.61 35.33 84.94 Delayed 4.53 3.54 3.75 5.51 6.30 11.81 Terminated 0.20 0.69 1.57 1.38 1.08 2.46 Not Started 0.20 0.20 0.39 0.59 0.20 0.79 Total 29.92 30.61 39.47 57.09 42.91 100 Pearson 2ð Þ ¼6 13:056 ðp-value : 0:042Þ LR 2ð Þ ¼6 14:0112 ðp-value : 0:030Þ Pearson 2ð Þ ¼3 6:9869 ðp-value : 0:072Þ LR 2ð Þ ¼3 6:9879 ðp-value : 0:072Þ CIS3 No or low effect 38.20 17.06 12.04 39.49 27.82 67.31 Medium or high effect 19.59 7.76 5.35 18.83 13.87 32.69 Total 57.79 24.82 17.39 58.31 41.69 100 Pearson 2ð Þ ¼2 6:5578 ðp-value : 0:038Þ LR 2ð Þ ¼2 6:5771 ðp-value : 0:037Þ Pearson 2ð Þ ¼1 0:7741 ðp-value : 0:379Þ LR 2ð Þ ¼1 0:7734 ðp-value : 0:379Þ

(12)

of sample representation, 34% of small firms, 31% of medium firms, and 30% of large firms in the sample considered the effects to be medium or high. This indicates some small disadvantage to smaller firms.

In terms of technology levels, out of the almost one third of the sample firms who report that technology was of medium or high effect 19% were in the low tech sector and 14% in the high tech sector and thus firms reporting the effect are more likely to be in the low tech sector. Once again however, if we take account of sample representation, then a low tech firm has a probability of 32% of reporting a medium or high effect and a high tech firm has a probability of 33% which are hardly dramatic differences.

Overall, the Pearson’s 2 and the LR test statistics for the hypotheses that the rows and columns in the two-way tables are independent suggest that differences in the outcomes across various size classes are significant in CIS 2 and CIS 3, but classifying according to the level of technology we can accept the hypothesis the distributions are different (at a 10% significance level) for the CIS2, but both test statistics reject the null hypotheses for the CIS3. Thus, using the raw data, there is some support for the hypotheses that smaller firms suffer more financial constraints and that high tech firms are more constrained but in both cases one cannot interpret the evidence as strong.

4. The importance of the availability of finance

by firm size and industrial sector: an ordinal

logistic regression model

In Section 3 we have shown that in the raw data from two large samples of UK firms, a considerable proportion of establishments report that financial factors have hampered their innovative behaviour. However, classification by firm size and level of technology were less informative. In order to explore further whether firms in high-tech sectors are more financially constrained in their innovative activity, and also to investigate more formally whether enterprises in certain firm size classes are significantly more likely to report financial constraints than others, we have fitted ordinal logistic regression models separately to the CIS2 and CIS3 data sets. The different survey designs and different data availability require the different model specifications. For CIS2 we assume that the ordinal scale of the four category outcome, is ‘no hampered projects’, ‘projects seriously delayed’, ‘projects termi-nated after being started’, ‘projects not even started’. In particular, the four follow-ing outcomes are given the values 1 to 4 respectively if, on the grounds of the availability of finance: the firm reported no hampered projects; a project was seriously delayed; a project was terminated after being started; a project was not even started. For the CIS3 we fitted a two category ordinal logit model with ‘no or low’ effect of the availability of finance on innovation versus ‘medium or high’ effect. In this case the response variable simply assumes value 0 if the firm reported ‘no effects’ or ‘low’ effects, 1 otherwise.

(13)

Independent variables are one/zero variables representing firm size class (small or medium) and industrial sector, except for CIS2 where due to the fact that there are very few observations in some service sectors (with many zero cells) only some of the service sector dummies have been included in the estimated equation. For CIS3 most of the service sectors have been included. We use large firms in the ‘Manufacture of basic metals and fabricated metal product’ (a rather traditional sector) as the baseline throughout. Although it would have been possible to have a single high tech/low tech dummy there is sufficient data to enable industries to be individually represented and thus allow a more detailed post estimation evalu-ation of results on the technology dimension.

Descriptive statistics for the CIS2 and CIS3 data sets are presented in Table 4. Although firm size class is represented by one/zero dummies in the analysis, for the purposes of conveying characteristics of the sample firms, in Table 4 we include the number of employees in small medium and large firms (the numbers of firms falling in to each category is detailed in the notes to Tables 1 and 2).

Table 4 Descriptive statistics for the CIS2 and CIS3

CIS2

... ...CIS3

Mean Std. Dev Mean Std. Dev

Firm class Small 25.61 11.86 21.81 10.98 Medium 120.69 52.76 106 51.45 Large 1134.13 2267.5 873.54 1470.41 Industrial sector Food 0.074 0.261 0.033 0.179 Textiles 0.065 0.247 0.026 0.160 Wood 0.089 0.286 0.066 0.249 Chemicals 0.080 0.271 0.018 0.132 Rubber 0.087 0.283 0.320 0.175 Office machinery 0.086 0.280 0.600 0.237 Basic metals 0.110 0.313 0.330 0.179 Electrical machinery 0.209 0.040 0.700 0.254 Transport equipment 0.093 0.291 0.040 0.203 Machinery n.e.c 0.043 0.199 0.057 0.232 Electricity 0.026 0.158 0.0066 0.078 Construction 0.025 0.155 0.0109 0.312 Sales 0.040 0.063 0.123 0.329 Trade 0.010 0.031 0.080 0.272 Transport – – 0.014 0.115 Telecommunication – – 0.049 0.216 Financial intermediation – – 0.035 0.185 Computer – – 0.022 0.146 R&D – – 0.0066 0.081 Engineering – – 0.0321 0.176

Sample sizes: CIS2:1016, CIS3:7240.

(14)

Table 5 reports the estimated coefficients together with the estimated standard errors (given in parentheses). An asterisk indicates if (under the null hypothesis that the estimated coefficient is equal to zero) according to the Wald statistic a given covariate significantly affects the probability of reporting that availability of finance was a constraint. For ease of interpretation we also report the odds Table 5 Results of fitting an ordinal logistic model to CIS2 and CIS3 data

CIS2

... ...CIS3 Estimated coefficient Odds ratios Estimated coefficient Odds ratios

Small 0:17 0:08 ð Þ** 1.186 0:25ð0:07Þ** 1.28 Medium 0:005 0:08 ð Þ 0.995 0:05ð0:08Þ 1.05 Food 0:123 0:28 ð Þ 0.885 ð0:680:142Þ** 1.97 Textiles 0:27 0:21 ð Þ 1.309 0:59ð0:16Þ 1.79 Wood 0:94 0:25 ð Þ** 0.391 0:34ð0:11Þ** 1.40 Chemicals 0:37 0:16 ð Þ** 1.450 0:54ð0:19Þ** 1.72 Rubber 0:29 0:15 ð Þ** 1.341 0:46ð0:15Þ** 1.58 Office machinery 0:17 0:14 ð Þ 1.180 0:69ð0:14Þ** 1.99 Electrical machinery 0:54 0:11 ð Þ** 1.718 ð0:50:10Þ** 1.65 Transport equipment 0:25 0:13 ð Þ** 1.281 0:43ð0:13Þ** 1.54 Machinery n.e.c 0:08 0:16 ð Þ 1.088 ð0:60:11Þ** 1.79 Electricity 1:29 0:31 ð Þ** 0.274 0:38ð0:32Þ 1.46 Construction 0:72 0:23 ð Þ** 0.486 0:22ð0:09Þ** 1.25 Trade 0:76 0:32 ð Þ** 2.145 – -Transport 2:89 0:41 ð Þ** 18.09 0:22ð0:10Þ** 1.25 Telecommunication – – 0:50 0:22 ð Þ 1.66 Financial intermediation – – 0:22 0:14 ð Þ* 0.80 Computer – – 0:77 0:17 ð Þ** 2.17 R&D – – 1:4 0:3 ð Þ** 3.98 Engineering – – 0:35 0:15 ð Þ** 1.43 c1 1:96 0:10 ð Þ 1:16ð0:08Þ c2 3:82 0:12 ð Þ – c3 5:47 0:18 ð Þ – LRtest ½p-value: 2ð15Þ ¼ 180:04 ðP > 2¼0:000Þ 2ð19Þ ¼ 124:02 ðP > 2¼0:000Þ

Notes: **Significant at 5%. *Significant at 10%. Sample sizes: CIS2:1016, CIS3:7240. Note: for CIS3 ‘Trade’ has been excluded from the reduced form of the estimated model because it is highly non-significant.

(15)

ratios in the third column for the CIS2 and last column for the CIS3, respectively. These are calculated by taking the exponential of the estimated coefficients. The constants c1, c2, c3refer to the estimated cut off points for the probability of the four

possible outcomes. At the bottom of columns two and four the likelihood ratio test for the overall significance of the estimated coefficients is also reported.

In Table 5, an odds ratio below (above) one, indicates that for that particular covariate the probability of experiencing a more severe impact from a lack of avail-ability of finance is lower (higher) than the covariate in the baseline, while an odds ratio equal to one indicates that probability is the same as the covariate in the baseline. For example, the odds ratio of 1.186 indicates that for small firms the probability of a more severe impact from a lack of finance is approximately 19% higher than for large firms.

The estimation of the logit model confirms that in both CIS2 and CIS3, small firms were significantly more likely to suffer a greater impact from financial constraints than large and medium sized firms. The analysis by level of technology also reveals that there are different impacts across sectors. In CIS2 the identified high-tech sectors mainly show high odds ratios, e.g. office machinery, electrical machinery, transport equipment (motor vehicles) and machinery n.e.c whereas non high-tech sectors such as food, wood, electricity and construction tend to show lower ratios (the high estimated coefficient for transport is due to the fact that there are only three firms in that particular sector and all of them reported an unsuc-cessful project). In CIS3 of the high-tech sectors identified, computers, R&D, office machinery, telecommunications, transport equipment, electrical machinery, and machinery n.e.c all have odds ratios greater than unity and greater than in some of the low-tech sectors such as construction and transport. The most obvious difference is between the low tech and high tech service sectors with average odds ratios of 1.32 and 2.26 respectively.

The goodness of fit of the estimated logit models has been assessed by comparing the observed and the predicted probability of each outcome. For the CIS2, for each firm size class (i = 1,2,3), the ordered-logit predictions are the probability that P0

i, kþui, k lies between k  1 and k (for k = 1, . . . , 4). Thus,

PrðP0i, k¼0jSizeiÞ ¼PrðPi, k0 þui, k6c0jSizeiÞ ¼

1

1 þ expðc0þP0i, kÞ

: The probability of outcomes 1 and 2 is given by

PrðP0 i, j, k¼jSizeiÞ ¼Prðc1<Pi, k0 þui, kcjSizeiÞ ¼ 1 1 þ expðcþPi, k0 Þ  1 1 þ expðc1þP0i, kÞ , for  = 1, 2, and PrðPi, k0 ¼3jSizeiÞ ¼Prðc3<P0i, kþui, kjSizeiÞ ¼1  1 1 þ expðc3þP0i, kÞ :

(16)

While for the model using the CIS3 we simply have PrðP0 i, k ¼0jSizeiÞ ¼PrðP0i, kþui, kc0jSizeiÞ ¼ 1 1 þ expðc0þPi, k0 Þ and PrðPi, k0 ¼1jSizeiÞ ¼PrðP0i, kþui, k>c0jSizeiÞ ¼1  1 1 þ expðc0þP0i, kÞ : In Table 6 we report the observed and the predicted probability of each outcome.

From Table 6 we can infer that the estimated models fit the datasets quite well since the predicted probabilities are remarkably close to the observed ones (the predicted probabilities being calculated at the mean of P0

i, k). Note that the

standard errors are reported in parentheses.

Table 6 also confirms that the findings re the importance of financial constraints differ between CIS2 and CIS3. We consider that this is at least partly because, although the questions in the CIS2 and CIS3 were similar (see Fig. 1), the percep-tion of the survey respondents to the quespercep-tions may have been very different. Table 6 Probability of the impact of finance availability conditional on firm size

CIS2

... ...CIS3

Observed Predicted Observed Predicted

ProbfP0 i, kjSmallg 0 0.835 0:837 0:048 ð Þ 0.660 0:661ð0:062Þ 1 0.151 0:133 0:039 ð Þ 0.340 0:339ð0:062Þ 2 0.007 0:024 0:008 ð Þ – – 3 0.007 0:006 0:002 ð Þ – – ProbfP0 i, kjMediumg 0 0.855 0:850 0:054 ð Þ 0.690 0:687ð0:058Þ 1 0.116 0:122 0:039 ð Þ 0.310 0:313ð0:058Þ 2 0.022 0:022 0:013 ð Þ – – 3 0.006 0:0066 0:004 ð Þ – – ProbfP0 i, kjLargeg 0 0.855 0:862 0:042 ð Þ 0.692 0:693ð0:061Þ 1 0.095 0:113 0:034 ð Þ 0.308 0:307ð0:061Þ 2 0.040 0:020 0:007 ð Þ – – 3 0.010 0:005 0:002 ð Þ – –

Notes: CIS2: 0 = no hampered projects, 1 = Seriously delayed, 2 = Terminated prematurely, 3 = Not even started. CIS3: 0 = ‘No’ or ‘Low’ effect, 1 = ‘Medium’ or ‘High’ effect.

(17)

In our view, the disparity of responses is embedded in the different nature of the questions asked. Even so, for CIS2 the results indicate that the (predicted) prob-ability of a project not being hampered by financial factors is slightly lower for smaller firms, with small firms tending also to suffer more serious effects when the project is constrained. For CIS3 ‘no or low’ effect is predicted as more common for medium and larger sized firms and a ‘medium or high’ effect as more common for small firms. Jointly, even with the expected in built downward biases, the results thus indicate that smaller firms are more affected than medium or larger firms.

5. Conclusions

In this paper we have used data from the second and third Community Innovation Surveys as conducted in the UK to explore whether financial factors constrain innovation and whether the importance of such constraints varies across firm sizes and sectors. Initial explorations of the different data sets available to us indi-cate that financial factors, specifically factors relating to either the cost of finance or the availability of finance, rank among the more significant factors that have acted as hindrances to innovation both in 1994–96 and 1998–2000.

Analysis of the CIS2 data (individual returns for UK firms, 1994–1996) indicates that (correcting for firm size) there is evidence that a firm in a high-tech sector has more chance of experiencing a greater impact from financial constraints than a firm in a low-tech sector.

Further results using the CIS2 data also provides clear evidence that once one has corrected for industrial sector then small firms experience greater risks of experiencing a greater impact from financial constraints than do large firms. The CIS3 data set (individual returns for UK firms 1998–2000) confirms these results. Overall one may thus conclude that financial factors do impact upon innovative activity. That impact is more severe in higher tech sectors and for smaller firms. These results are in line with existing results in the literature. They arise however from using a different methodology to the cash flow method that is more standard. The two methodologies have their own costs and benefits but in our view should be considered as complementary rather than competing. The benefits of the current methodology is that, relative to existing results, it encompasses a wider class of phenomena than is usually studied (innovation rather than R&D) uses a different data set than other studies and also helps to overcome conceptual problems with the use of a cash flow covariate.

Acknowledgements

This work has been financed under the Fifth EU Framework Programme, Key Action: Improving the Socio-economic Knowledge Base, Contract No. HPSE-CT-1999–00039. We would like to thank two referees of this journal for comments on an earlier draft. Thanks are also due to participants in the MSM seminar at Warwick Business School. All errors that remain are the responsibility of the authors alone.

(18)

References

Basu, P. and Guariglia, A. (2002) Liquidity constraints and firms’ investment return behav-iour, Economica, 69, 563–82.

Bond, S. and Cummins, J. (2001) Noisy share prices and the q model of investment, Discussion Paper No. 22, Institute for Fiscal Studies, London.

Bond, S., Klemm, A., Newton-Smith, R., Syed, M., and Vlieghe, G. (2004) The roles of expected profitability, Tobin’s Q and cash flow in econometric models of company invest-ment, Working Paper No. 04/12, Institute for Fiscal Studies, London.

Bond, S., Elston, J. A., Mairesse, J., and Mulkay, B. (2003) Financial factors and investment in Belgium, France, Germany and the United Kingdom: a comparison using company panel data, The Review of Economics and Statistics, 85, 153–65.

Carpenter, R. and Petersen, B.C. (2002a) Is the growth of small firms constrained by internal finance? The Review of Economics and Statistics, 84, 298–309.

Carpenter, R. and Petersen, B.C. (2002b) Capital market imperfections, high tech invest-ments and new equity financing, Economic Journal, 112, F54–72.

Cummins, J., Hasset, K., and Oliner, S. (2006) Investment behavior, observable expecta-tions, and internal funds, American Economic Review, 96, 796–810.

Hall, B. (2002) The financing of research and development, Oxford Review of Economic Policy, 18, 35–51.

Hatzichronoglou, T. (1997) Revision of the high-technology sector and product classifica-tion, STI Working Papers (97) 216, OECD, Paris.

Hubbard, R.G. (1998) Capital market imperfections and investment, Journal of Economic Literature, 36, 193–225.

Kaplan, S. and Zingales, L. (1997) Do investment cash flow sensitivities provide useful measures of financing constraints? Quarterly Journal of Economics, 112, 169–215.

Stiglitz, J. and Weiss, A. (1981) Credit rationing in markets with imperfect information, American Economic Review, 71, 393–410.

Young, A. (1996) Measuring R&D in the services, STI Working Papers (96) 32, OECD, Paris.

Appendix

UK CIS survey details

The Community Innovation Survey has been undertaken three times (although as sample selection and other factors were suspect for CIS1 this is little used), design and application making the resulting data suitable for comparison across countries. The survey is designed on the basis of a harmonised (across the European Community) questionnaire, although the questionnaire has changed over time, with individual countries on occasion adding extra questions to the basic survey instrument. The questionnaire has been applied to large random samples of firms in each country, national statistical offices being largely responsible for conducting the surveys. Details of the CIS survey and questionnaires etc can be found at www.cordis.lu/innovation-smes/src/cis.htm.

(19)

CIS2 was conducted by the UK Office for National Statistics on behalf of the Department of Trade and Industry. The reference period was 1994–96, response was voluntary and the survey was conducted by means of a postal questionnaire. The sample was drawn from enterprises in the UK with more than 10 employees in sections C–K of the Standard Industrial Classification. A stratified random sample was used with a minimum of five enterprises selected in each stratum. The sample was stratified by SIC 92 two digit class and eight employment size bands. The sampling frame was the Inter Departmental Business Register of 155,000 companies of whom 5892 were sampled and 2344 final responses were received (a 43% response rate after excluding enterprises which had ceased trading). The questionnaire (12 pages) covers various aspects of firm innovative perform-ance, other firm characteristics and also factors influencing innovative activity. Table A1 Classification of economic activities

Manufacturing Low-technology

Food Manufacture of food products; beverages and tobacco. Textiles Manufacture of textiles and textile products;

Manufacture of leather and leather products.

Wood Manufacture of wood products; manufacture of pulp, paper and paper products.

Rubber Manufacture of rubber and plastic products; Manufacture of other non-metallic mineral products. Basic metals Manufacture of basic metals, and fabricated metal products. Chemicals Manufacture of coke, refined petroleum products, and nuclear

fuel,

Manufacture of chemicals. High-technology

Transport equipment Manufacture of transport equipment.

Office machinery Manufacture of office machinery and computers. Machinery n.e.c. Manufacture of machinery and equipment not elsewhere

classified.

Electrical machinery Manufacture of electrical and optical equipment. Services

Low-technology

Construction Construction.

Electricity Electricity; gas and water supply.

Wholesale and retail trade Wholesale and retail trade, repair motor vehicles, motorcycles and personal and household goods.

Transport Land transport; transport via pipeline; water transport; air transport.

High-technology

Telecommunications Telecommunications. Financial intermediation Financial intermediation. R&D Research and development.

Computers Computers and engineering activities, and related technical consultancy.

Engineering Architectural, engineering activities, and related technical consultancy

(20)

CIS3 was also conducted as a voluntary postal questionnaire by the UK Office for National Statistics, the reference period being 1998–2000. The questions were similar to, although in some cases variations upon, the CIS2 questions. A copy of the questionnaire can be found on www.dti.gov.uk/tese/science.htm. Compared to CIS2 the sample was extended to include agriculture; fishing and forestry; public administration and defence; education; and health and social work. The survey again covered enterprises with 10 or more employees in sections C–K of the industrial classification. The survey was stratified by region, by SIC groupings and by five employment size bands. Almost 16% of the 126,775 enterprises in the population were sampled (13,340) in April 2001. A top up survey of England only (another 6287 enterprises) to provide greater regional data was undertaken in November. Of the 19602 enterprises selected 8,172 responses were received (a 42% response rate). On average each respondent represents 23 enterprises in the population.

Riferimenti

Documenti correlati

Table 8: Fixed Effects estimation of housing investment decision Dependent variable: housing weight – fraction of net housing wealth over total net wealth All columns report

We deal with potential endogeneity with a fixed effect instrumental variable approach, using as instruments the presence of at least one economics graduate in the

Nella loro varietà, tuttavia, i singoli argomenti ricompongono attraverso il tempo un panorama esemplificativo dei vari tipi di esposizioni d’arte medievale, da quelle più

Come sotto- lineato dall’autore nelle conclusioni del volume, occorre un passo fondamentale, ossia riconoscere agli studi quantitativi della scienza «lo status di scienza sociale

Gli arabarchi, in pratica, gestivano tutto il sistema fiscale che ruotava attorno al commercio con l’Oriente, incaricandosi di sorvegliare le merci lungo il cammi- no nel

Dice Domenica Cinti parlando di Battista Mazzoni: “Lui veniva quando voleva perché essendo mio sposo non se li vietava di venire” (Sarti, 1999, p. Cosa c’è di strano in

Nel punto esatto in cui servono i migliori tacos di gambero (secondo il mio punto di vista) della Zona Centro di Tijuana, convergono vagabondi che chiedono soldi per un taco che

Using the High Energy Spectroscopic System (H.E.S.S.) telescopes we have discovered a steady and extended very high-energy (VHE) γ-ray source towards the luminous blue