Financial innovation in the UK

Abstract

This study employs a national survey of over 1100 British financial firms to ascertain the determinants of financial innovation and their sales success using the logit and the generalized Tobit models. We find that the likelihood of financial innovation rises with the size of financial firms, employee education, greater expenditure on research and development, the availability of finance and the extent to which firms cooperate with each other. Perceptions of economic risk and innovation costs are also influential. R&D, cooperation and human capital are the main variables driving the success of financial innovation, measured by the percentage share of innovations sold. Firms in London/the south have a significantly greater tendency to innovate, though Scotland also does well. Stock broking, fund management and related activities are more innovative than firms in the financial intermediation and pension/insurance sectors.

Keywords:

• determinants/success of financial innovation
• ISCORE
• UK financial sector

JEL Classification::

• G1
• G19
• G21
• G22
• G23

Acknowledgements

The author is grateful for funding from the University of Macau. We greatly appreciate the feedback from Peter Sinclair, Stefan Hunt, Tony Saunders, one anonymous referee, and the Department of Innovation, Universities, and Skills (Marion Frenz, Ray Lambert and Stephanie Robson), which also supplied the survey data. Comments from the Woltpertinger, Midwest Finance Association and Federal Reserve Bank of New York/CEPR/Bank of Finland conferences proved very helpful, especially those from Ted Gardiner, Phil Molyneux and Javier Quesada. All errors are our responsibility.

Notes

¹ The list is not exhaustive but the US leads in just two areas, corporate finance and exchange-traded derivatives (Heffernan, 2005, table 2.2).

² Surveys can be found in Cohen and Levin (1989), Cohen (1995), Molyneux and Shamroukh (1999) and Unger (2005).

³ The indirect academic link is measured by the average number of each banks’ employees on the editorial or advisory boards of two academic/practitioner journals.

⁴ A copy of the questionnaire can be found on the website http://www.berr.gov.uk/dius/innovation/innovation-statistics/cis/cis4-qst/page11578.html

⁵ Formerly the Department of Trade and Industry or DTI. In 2007, its functions were transferred to the Department for Business, Enterprise and Regulatory Reform (BERR) and the Department for Innovation, Universities and Skills (DIUS).

⁶ It is because firms fill in multiple years of data retrospectively.

⁷ Source: www.statistics.gov.uk/methods_quality/sic/downloads/UK_SIC_Vol1(2003).pdf

⁸ Source: www.berr.gov.uk/innovation/innovation-statistics/cis/page10957.html

⁹ In the manufacturing literature, this is known as ‘diffusion innovation’. See Dosi (1988), Freeman (1994), Ray (1984) and Reinganum (1989), among others.

¹⁰ The survey questions do not explicitly state that the good/service had to be a relatively new innovation. So it is possible that firms giving an affirmative answer could have decided to offer a product that had been on the market for years. However, given the survey was about financial innovation (which dates quickly) over a specified period, most firms probably interpreted the question as offering goods/services recently launched on the market by innovators. After consultation with the DIUS we interpreted a positive answer to be imitators selling relatively new innovations already on offer by competitors. The questionnaire allowed firms to report sales of innovations that were either new-to-the-market or new-to-the-firm; they were treated as mutually exclusive.

¹¹ See Cohen and Levin (1989) for more details.

¹² Using Equation 2 the ISCORE for each firm is calculated, each is grouped according to relevant sectors/regions, and an average value for each sector/region is obtained. The equivalent of the ISCORE was developed by Cosh et al. (2004).

¹³ See Table 4.

¹⁴ The propensity score-matching model is used to correct the sample selection bias derived from observable differences between the treatment and comparison groups (Dehejia and Wahba, 2002). Greene (2003) provides a good example of treatment effects. Suppose earnings are used to measure the value of a university education because the researcher cannot divide a group into whether or not they were educated to university standard. Here, a potential problem with self-selection exists: an individual might enjoy higher earnings independent of whether or not he/she attended university. Our sample does not suffer from this problem. However, SI is only observed if SI+ = 1, (a different type of sample selection bias) best addressed through a generalized Tobit model.

¹⁵ No information on firm-specific internal funding is available.

¹⁶ Innovation is a dynamic process. Hence, investments in R&D and employee education may pay off much later. Given the data set only contains information on the UK innovation over 3 years, one cannot be certain that the dependent variable adequately captures the innovation that may be caused by the independent variables. Thus, the empirical results shall be interpreted carefully.

¹⁷ In Parisi et al. (2006), the percentage of correct classification ranges between 64% and 73%.

¹⁸ Though SIZE and log(SIZE) were found to be highly correlated with each other, both were included in the estimations because of the importance of testing for whether mid-sized firms are significant innovators. The significant correlation is not of concern because the size coefficient is insignificant and estimations without SIZE (not reported but available on request) did not alter the results.

¹⁹ As indicated in Bertrand and Schoar (2003), a great extent of the heterogeneity in investment, financial and organizational practices of firms can be explained by the presence of manager-fixed effects. For example, certain CEOs might increase both R&D expenditure and financial innovation. However, data on CEO characteristics are not available. Thus, the estimates might suffer from omitted variable bias.

²⁰ The firms are coded in the survey, so it is not possible to match firm-specific accounting data to test for a link between profit and innovation, among other things.

²¹ The results of the test are available from the authors upon request.

²² The only difference from Equation 4 is that for SI+, SI+_NF and SI+_NM, the log(SIZE) coefficient is positive and significant at the 1% level, whereas this variable is only significant (at the 10% level) for the coefficient on SI_NM.

²³ In a comprehensive review article by David et al. (2000), one-third of the studies find that public R&D funding behaves as a substitute for private funding (at firm level), rising to just under 50% for studies conducted at industry or higher levels of aggregation. Hall and Van Reenen (2000) review the evidence on fiscal incentives for R&D across OECD countries and conclude that $1 in tax credit results in $1 of additional R&D. Nguyen (2007), in a study using the CIS survey data for Luxembourg, concludes that innovative firms in the manufacturing sectors receive more public funding compared to the service sector. Independent of the sector, large firms appear to benefit more from public subsidies.

²⁴ This does not contradict the earlier finding which showed no sector or region was a factor affecting the likelihood of FI.

²⁵ For the purposes of this discussion, the number of regions is reduced to nine (in line with chart 1) because the scores for certain adjacent regions were so similar.

²⁶ London and the South may always score highest because of the number of financial firms located in the City as well as headquarter effects. For example, the innovation may be made by workers in Leeds but the benefit is for the entire organization and recorded by an individual at headquarters.