BUREAU OF ECONOMIC RESEARCH 1050 Massachusetts Avenue Cambridge, MA 02138 December 2008 We would like to thank the Mediocredito-Capitalia research department for having kindly supplied firm level data for this project.
We then apply the model to data on Italian SMES from the"Survey on Manufacturing Firms"conducted by Mediocredito-Capitalia covering the period 1995-2003.
According to the latest available data from the Census, more than 99 per cent of active firms (out of 4 million) have fewer than 250 employees (95 per cent have fewer than 10 employees,
along with a description of the data used in this analysis; Section 4 concludes with a discussion of the results
The model is designed specifically to work well with innovation survey data, from which it is possible to directly measure other aspects of innovation in addition to R&d expenditures.
Given the increased diffusion of this type of micro data across countries and among scholars, many empirical explorations of the impact of innovation on productivity have relied on the CDM framework. 2 In particular
See Section 3 of this paper for further information on the data. 6 CDM model specification allowing our model to separate the impact of different kinds of innovation (product
and process) on firms'productivity. 3. Data and Methodology The data we use come from the 7th,
We merged the data from these three surveys excluding firms with incomplete information or with extreme observations for the variables of interest. 4 We focus on SMES,
Their model-specifically tailored for innovation survey data and built to take into account the econometric issues that arise in this context-is made up by three blocks,
since the effect of R&d on productivity can vary a lot with the technological content of an industry (see Verspagen, 1995 for a cross country, cross sector study and, more recently, an analysis based on micro data by Potters et al, 2008).
Because of the way our data and innovation survey data in general is collected, the analysis here is essentially cross-sectional.
In addition, the innovation data are collected retrospectively (innovating over the past three years), and the income statement data is mostly contemporaneous.
As a robustness check we estimated the same 3 equation model using R&d intensity lagged one year instead of contemporaneous R&d intensity
and Baldwin and Gu, 2004, for an exploration using Canadian data), and this effect is particularly strong for high-tech firms,
However, in our data we also have a measure of capital available, constructed from investment using the usual declining balance method with a depreciation rate of 5 per cent
we built building a slightly different sample of firms from our data that removed firms with fewer than 20 employees
Table 6 shows results from Griffith et al. 2006 for the four countries and for a variation of our model applied to these data for Italy. 14 The last column
Thus it appears to be difficult to find strong evidence of innovation underperformance in these data,
and we hope to explore the question further in the future using these data. Acknowledgements We would like to thank the Mediocredito-Capitalia (now Unicredit) research department for having kindly supplied firm level data for this project.
We thank also Susanto Basu, Ernie Berndt, Piergiuseppe Morone, Stéphane Robin, Mike Scherer, Enrico Santarelli
a Sensitivity analysis, Economics of Innovation and New Technology. Vol. 15 (4/5), pp. 317-344.
New Evidence using Linked R&d-LRD Data, Economic Inquiry, Vol. 29 (2), pp. 203-228.
a Germany-Italy Comparison Using the CIS Database, Empirica, Vol. 28, pp. 293-317. Lööf, H,
A Reassessment Using French Survey Data, The Journal of Technology Transfer, special issue in memory of Edwin Mansfield, Vol. 30 (1-2), pp. 183-197.
Testing Sectoral Peculiarities using Micro Data, IZA Discussion Paper N. 3338. Rajan, R. G, . and L. Zingales (2003), Banks and Markets:
Data are from the third Community Innovation Survey (CIS 3) for France, Germany, Spain, and the U k. Results for Italy come from Tables 3-5 of this paper.
a) This column shows data for all 3 periods in Italy (1995-1997,1998-2000,2001-2003). 29 Figure 1 Value added per employee.
(Census data)% of firms with innovation (CIS survey on firms with more than 10 employees) 31 Appendix Variable Definitions R&d engagement:
Data are from the third Community Innovation Survey (CIS 3) for France, Germany, Spain, and the UK.
Data for Italy are from the Mediocredito Surveys. Among the several variables included in the original table,
we selected only those comparable to our data. Data are weighted not population. a) This figure encompasses all the subsidies, regardless their source.
b) This column shows data for all 3 periods in Italy (1995-1997,1998-2000,2001-2003.
Units are logs of euros (2000) per employee. 34 Table A2 A nonparametric selectivity test Dependent variable Prob (R&d>0) R&d expend. per employee D (Large firms
and proprietary databases to help customers use their products more effectively. Changes in organization structure and culture are required almost always to do this effectively.
Many SMES don't recognize the value of data, have minimal archives and don't learn from experience (Woodcock et al.,
The OEM collects data at the sale and has reduced costs associated with acquiring new customers.
Additionally the data accumulated from having a greater knowledge of customers'behavior enables the company to continually add value
The importance of acquiring and building databases as tools for adding value and defending against competitors is returned to later).
and its database of trip costing enables the company to accurately quote on trips and to provide customized and traceable service.
This requires GF's computer systems to seamlessly integrate with car plants exchanging data in real-time.
For example, using advanced data collection and data mining tools, coupled with real-time data collection over the Internet may provide a whole new level of product and service reliability.
The third mini-case provides an example. Mini-Case Example#3: Taprogge Gmbh,(www. taprogge. com) a family owned business headquartered in Germany,
The latest equipment has a number of embedded sensors that monitor the performance and relay the data over the Internet back to a central office.
Analysis of these data enable the company to predict possible performance deterioration and ship parts followed,
The company's most valuable asset is a complex database covering all operating parameters of every installation.
These data enable Taprogge to a) predict the behavior of a system in most if not all locations
The database analyzed was the ISI Web of Knowledge (ISI, 2006) using the social science sub-set of indexed publications.
The fall in the price of computers and data storage devices coupled with the rise of the Internet, have made the use of digital information as a competitive weapon no longer just the domain of larger companies.
Start-up companies can now harvest information technology to provide their customers with greater value and to create subtle barriers to competition.
Data Acquisition and Mining: Capturing data on customer requirements and using it to create unique services
or products can be a powerful way of adding value and keeping out competitors. Netflix has changed the way that consumers rent movies.
but the ability to mine the data obtained by combining information from ALL customers nationwide.
by getting instant feedback from their database (customers provide long lists of future wants and rate past rents),
something not 25 possible to do on a local basis. Using this novel database structure, Netflix is able to provide its customers with a convenient personalized service,
The following mini-case shows how, in a business-to-business market, acquisition of data, and its subsequent analysis or mining can provide a powerful service model for a manufacturer.
The proprietary data that the company collects on its clients'unique situations are a major competitive advantage,
A business model based on information sharing can provide high barriers against competitors as the costs involved in integrating compatible data
A sound business model using data lock in will have multiple partners so that the dependence on one partner is reduced.
Entirely new forms of business can be created by employing data acquisition and mining to lock in customers, suppliers and partners.
and coded database of proprietary cleaning formulas for specific customer needs, whether to clean eggpacking equipment or the floors of a car assembly plant.
In exchange, the franchisee gets access to the database on demand when a customer need is defined.
In the event that there is no solution in the database for a customer's new problem,
then the franchise agreement commits them to submit it to the central database, where it becomes available to all franchises
In this way, the franchise model is enhanced by the continual building of a proprietary database of customer solutions adding greater value to both the franchisor and franchisees.
and the new solution now becomes an integral part of the Chemstation database. The sharing of such information by the franchisees with the HQ is mandated by a written agreement between Chemstation and its franchisees.
The database is a key asset for Chemstation and it has the necessary software and framework in place to interpret the results
and distribute the data. The database also builds barriers against competition. For example, Chemstation solved a cleaning problem at a Harley davidson plant within its shock absorbers manufacturing division
which resulted in using one cleaning solution on one line and another solution for the adjacent sister line.
This subtle know-how becomes a part of Chemstation's data bank. Such captured knowledge helps to lock in customers
if there are quantitative preference data (e g.,, from surveys. Tools such as cluster analysis have been used successfully for this purpose.
Generate and Assemble Ideas. Once the target segment and its core needs are identified, the next task is to generate ideas to address these needs.
and it is critical to vet the input data carefully;(2) the business analysis should only serve as guidelines,
compared to the hard data (e g.,, projected market share, net present value. They are (1) strategy fit;(
I obtain data from many different sources; we listen to suggestions from suppliers; we use consultants in focused roles Support The degree to
and the firm's willingness to invest in data capture and storage. The move to Phase III should not be made until the firm has mastered thoroughly selling services with current sales
if captured, may help create a proprietary database that can give the firm a competitive advantage over its rivals.
and Taprogge used IT to capture data about their products'performance in different contexts and developed proprietary databases that allowed them to customize use of their product to meet specific customer needs.
All five cases required 51 their work force to be trained extensively in activities that their customers could not do as well as they could.
we analyzed the MEP database of past success stories coupled with selective interviews. We reviewed a total of 689 success stories from 2002 to 2005 that were posted on the NIST MEP website.
A further shortcoming is the static nature of the case histories that reside within the MEP database.
We found this database rather difficult to use. However, the more important issues are concerned with both content and format.
The current MEP database may not fulfill the purpose for which it was created, either in content or in user value.
Recommendation 1-Leverage the existing skills between MEP centers We recommend augmenting the current database with a dynamic knowledge network.
Recommendation 3-Analyze use of the current MEP database We recommend a web-survey of existing MEP centers to determine:
The current use and value of the database Research to determine features and expectations of the knowledge portal Market research for the proposed training courses and content Recommendation 4-Develop a research agenda As reported in section 3,
Within this category are the case histories that reside in the MEP database. They are rather sterile reports that are built around standard terms, categories and data,
and do not really convey any of the actual touchy-feely attributes which are much more important in this case.
Explicit-to-explicit is the simplest transfer challenge exemplified best by web-publishing such as the MEP database.
Tacit-to-explicit has been the Holy grail for many years spawning the field of expert systems. The aim has been to somehow capture the subtleties of tacit knowledge
Bios and contact information for the consultants. 59 The aim of these tools and support data is to prime the outreach function at the MEP offices on
and thereby become part of an ongoing and active database for future projects as the links can be used to tie in the experts for future cases. 60 7. 0 REFERENCES Agarwal, R. and J. Prasad (1997)."
Analysis of data from the U s. Bureau of Economic Analysis. A. Warren, Personal Correspondence. Ratio of 82.5 is taken at Q1 in 2005.
Clustering is particularly important to gain access to new ideas and tacit knowledge, especially in young industries.
For example, the Gellman (1976,1982) data base identified SMES as contributing 2. 45 times more innovations per employee than do large firms.
In a clustering strategy, firms take advantage of linkages with other enterprises afforded by geographic proximity
Data constraints can be overcome to study the extent of knowledge spillovers and their link to the geography of innovative activity using proxies like patenting activity, patent citations,
Therefore comparability of the data in this table is guaranteed not fully. 21 Year founded. 22 Not included:
and Han Zhang, 1999, Small Business in the Digital economy: Digital Company of the Future, paper presented at the conference, Understanding the Digital economy:
Data, Tools, and Research, Washington, D c.,25-26,may 1999. Berman, Eli, John Bound and Stephen Machin, 1997,‘Implications of Skill-Biased Technological Change:
International Evidence, working paper 6166, National Bureau of Economic Research (NBER), Cambridge, MA. Bessant, J.,1999, The Rise and Fall of Supernet:
OECD. OECD, 1999, Cluster analysis and Cluster-based Policy in OECD countries, Paris: OECD. Porter, M. 1990), The Comparative Advantage of Nations, New york:
Prevenzer, Martha, 1997,‘The Dynamics of Industrial Clustering in Biotechnology,'Small Business Economics, 9 (3), 255-271.
number n Number of points of data making the SIC no Data point number p n Number of periods of SIV analysis O Periodicity coefficient Periodicity compression coefficient
and ability to generalize from the produced data, have different requirements and limitations than in other disciplines.
That necessitated a discussion of the axiological and epistemological aspects of the research performed in this thesis. The initial work to build the desired model (Abouzeedan 2001) applied a textual/statistical analysis method to existing basic information from a Swedish database (Affärsdata
I preferred to utilize existing data without pre-structuring. Although I relied on existing accounting data for the financial parameters,
there were no predetermined requirements on how the data would be displayed. Finally, although the major outcome was an empirical model,
verbal descriptions and explanations (i e. narrative-textual analyses) were used in a number of papers that addressed the issue of performance in relation to the external environment of the firm,
and noisy data sets (Jain and Nag 1997). Furthermore, decisions based on financial failure prediction, which is driven statistically, may actually trigger a bankruptcy.
One of the problems with using the financial ratio approach to predict company performance is the huge number of such ratios that can be deducted from the available financial data for larger firms (Chen and Shimerda 1981.
who based his works on analyzed data collected by Roethlisberger and Dickson (1939). Social psychologists such as Likert (1961) and Katz et al.
The clustering I chose for the parameters in the intended model is based on the understanding that the parameters in each subset are interconnected closely.
where data are analyzed and interpreted (Brannen 2005). Traditionally quantitative methods are concentrated more on input issues.
The dominant method in my qualitative research approach was to compile data into review articles and conceptual papers.
That is why I saw, in the case study, a methodical approach to retrieve empirical data and to satisfy both forms of logic.
In quantitative research, observation is not generally 55 considered a very important method of data collection for two reasons.
Two of the papers (3 and 7) of this thesis used case study methods with textual analyses and analyses of accounting data.
Qualitative methods such as case studies allow for multiple data-collection methods under the same study, unlike quantitative research studies (Chetty 1996.
One of the best methods of collecting data is indepth interviews (Welch and Comer 1988.
Data can be analyzed using different techniques (Chetty 1996. The writer recommended using a single case study method in SME research.
The argumentation and reasoning carried in the paper is general in its nature (d). 2 Coverage Intensity
the data was taken directly from the accounting reports of the firm and the analysis was performed while
The technology intake data can be taken directly from the financial records or deducted from this information.
the data was delivered from the firm management for the period of the analysis; and I have good access to the situation of the firm.
their coverage intensity (the vertical axis), and their information intensity requirements (the horizontal axis). Both of these parameters are valid measures for the nature of the output generated from the different SME performance models.
In paper 3, the validity of the data used in the analysis of the firm stems from two facts:
60 and the data was taken directly from the accounting reports of the firm for the period of the analysis. Also,
In paper 7, the data used in the analysis of the firm is valid for three reasons:
the data was delivered from the firm's management for the period of the analysis; and the owner of the firm is a close friend of mine
and have defined its limits within a specific context determined by the data input. In the case of the SIV model
Such evaluation should utilize the existing data and complete it with more new data reflecting the additional years of analysis incorporated.
It is important to highlight that reliability should be understood in relation to the research method used in this case
The technology intake data can be taken directly from the financial records or deducted from the accountancy information
The technology intake data can be taken directly from the firm's financial records. In this particular case
and provided all the necessary input data. This secured the reliability of the analysis in paper 7. There are problems related to granting reliability of measurement in the papers of the thesis. Basically,
which included Z-Scores, ZETA Scores, and Neural networks (NN). The strengths and weaknesses of each model were exposed
and Neural networks are examples of models that relate to internal factors. Utilizing SMES indiscriminately will negatively affect the outcome of the majority of SME studies.
The vertical axis indicates the coverage intensity of the model, varying from an individual firm up to a whole group of firms.
The vertical axis indicates the coverage intensity of the model. The first parameter considers the information requirements of the models,
such as the ZETA and Neural networks models, require a high level of information intensity. That implies the need for detailed data,
which is something that SMES generally lack. The desired model requires a reasonably moderate data input to counter the issue of SMES'accounting
and reporting techniques, which provide less intensive information input than those of large firms. The SIV model has a moderate information level.
As the coverage intensity (the vertical axis) becomes lower the analysis becomes more focused on the internal environment of the firm.
Examples for such models are the ZETA model, the Neural networks model, and the SIV model.
When coverage intensity is high, the models have more focus on the external environment. Examples of those models include the stochastic models and Hazard modeling.
The learning model has a relatively intermediate level of coverage intensity indicating a dual focus. 71 Paper 3:
which require a larger number of business ratios, the SIV analysis uses basic accountancy data, without advanced statistical methods of variable elimination.
The other group includes Z-Scores, ZETA Scores, Neural networks, and the SIV model. These are more suitable to the investigation of firm performance in relation to the internal environment of an enterprise.
The vertical axis indicates the coverage intensity of the model. The first variable represents the information requirements of the models
such as the ZETA and Neural networks models, require a high information intensity level. Such a requirement can be a problem
and does not use sophisticated statistical methods to eliminate input data. Rather, it uses limited accountancy information in an efficient way.
the SIV analysis can use basic accountancy data and does need not advanced statistical methods. The fishery firm had no innovation or development activities,
The desired model should have a reasonably moderate data input to counter the issue of SMES accounting
In that sense, graphical statistics play an important role in the interpretation of the data output of the model.
The first dimension is the SME model coverage level known as coverage intensity. The dimension of intensity ranges from a group of firms to a single SME.
The coverage intensity is presented by the vertical axis. The second dimension is the information intensity requirements of the model, known as the information intensity requirement.
It indicates the level of information input required by the performance evaluation model. The information intensity requirement is presented by the horizontal axis in the diagram.
In Irene Johansson (ed.),the Uddevalla Symposium 2002 Anthology (Research Reports 03:1), Innovation, Entrepreneurship, Regional Development and Public Policy in the Emerging Digital economy, University of Trollhättan
Regression for longitudinal even data. Beverly hills, California: Sage Publications. Altman, E. I. 1968. Financial ratios, discriminant analysis and the prediction of corporate bankruptcy.
Comparisons using linear discriminant analysis and neural networks (the Italian experience. Journal of Banking and Finance 18 (3), 505 529.
Neural networks versus logistic regression in predicting bank failure. In R. P. Srivastava (ed.)Auditing Symposium. Vol:
Data mining with neural networks: Solving business problems from application development to decision support. Mcgraw-hill, Inc. Hightstown, New jersey, USA.
Neural networks and the mathematics of chaos an investigation of these methodologies as accurate predictors of corporate bankruptcy.
The First International Conference on Artificial intelligence Applications on Wall street (Proceedings. IEEE, 52 57. Cainelli, G.,Evangelista, R. and Savona, M. 2004.
Generalization with neural networks. Decision Support systems 11 (5), 527 545. Edvinsson, L. and Malone, M. S. 1997.
Forecasting small air carrier bankruptcies using a neural network approach. Journal of Financial Management and Analysis 13 (19), 44 49.
Accounting data and the prediction of business failure, the setting of priors and age of data.
Performance evaluation of neural network decision models. Journal of Management Information systems 14 (2), 201 216. Jaques, E. 1951.
Using artificial neural networks to pick stocks. Financial Analysts Journal 49 (4), 21 27. Kumar, M. S. 1985.
An empirical investigation of some data effects on the classification accuracy of probit, ID3 and neural networks.
An artificial neural network approach to predicting the outcome of Chapter 11 bankruptcy. The Journal of Business and Economic Studies 4 (1), 57 73.
and a realistically proportioned data set. Journal of Forecasting 19 (3), 219 230. Mcpherson, M. A. 1995.
Presented in The Uddevalla Symposium 2002, Innovation, Entrepreneurship, Regional Development and Public Policy in the Emerging Digital economy.
Interpreting qualitative data: Methods for analysing talk, text and interaction. London, UK: SGAE Publications Ltd.
Trading equity index futures with a neural network: A machine learning-enhanced trading strategy. The Journal of Portfolio Management 19 (1), 27 33.
Trist, E. I. 1981. The evolution of sociotechnical systems as a conceptual framework and as an action research program.
Testing Gibrat's law with establishment data for Lower saxony, 1978 1989. Small Business Economics 4 (2), 125 131.
He may use retrospective data, but these bring little certainty since nobody is using them the way he suggests.
The authors are entirely responsible for the facts and accuracy of the data presented. 1 Foreword:
Project fact-sheets drafted with data based on interviews and desk research (one per project analysed) Telephone interviews with project lead partners
in particular, through analytical studies and EU-wide data and statistics. The overall objective of the programme is to foster a business-friendly environment for SMES with a view to ensuring
way (Figure 5). This figure particularly shows that the INTERREG IVC projects that have tackled objectives related to the theme of‘Innovation Capacity of SMES'offer a good and uniform coverage of the barriers identified.
Breakdown of sub-projects, GPS and other measures in terms of barriers addressed The bar chart shows that the seven projects under analysis offer a complete coverage of the most relevant barriers to the Innovation Capacity of SMES.
When this coverage is compared with the initial coverage in terms of the original objectives of the projects (Figure 5,
But for a project such as INNOMOT, this wider coverage in terms of barriers addressed by the Good Practices identified
The students are selected annually via a database of at least 350 students from all over the world (mostly Swedes).
while several programmes and measures exist in many forms across Europe to assist SMES with the transition towards a digital economy,
and cooperation Cluster management MINI-EUROPE Cluster Support Environment Model Clustering physical infrastructure requirements to facilitate growth
and internationalisation(§3. 2. 5). Medium Cluster policies SMART+SMESGONET Clustering management activities supporting the internationalisation
The INTERREG IVC website has a GP database, which is useful for an initial benchmark,
Partners in the analysed projects would like to see a more sophisticated capitalisation tool with a regularly updated database
a user searching for GPS would not have to look into the INTERREG database and the URBACT database,
he would search one global database). ESPON65: The European Observation Network for Territorial Development and Cohesion aims to support policymakers by providing territorial evidence as well as support.
and represent the demand for data to support policy development. Therefore, these projects are not about GPS,
but about data and case studies. Specific knowledge available from ESPON can help managing authorities including regional authorities to improve their policies.
INTERREG IVC project partners could include these data when defining their work programme, identifying GPS and analysing their conditions of transferability.
These three networking programme have a wealth of data relevant to regional policy improvement, especially for URBACT II and ESPON;
A capitalisation tool allowing easy access to these data would be beneficial to the future INTERREG EUROPE project partners.
a capitalisation tool including an up-to-date database and personnel to provide professional advice could include data from these networks.
Another way to improve synergies would be for the programme to require a benchmark analysis of the GPS that exist
a region wishing to import an innovation voucher scheme from one of the project partners would benefit from analysing all the different innovation voucher schemes in the database. 3. 4. 3 Synergies with other European Funds
The seven INTERREG IVC projects analysed offered a complete coverage of these barriers both in terms of initial objectives and activities actually carried out by the projects.
including those based on massive volumes of data or processing, to SMES with limited resources. 73 DISTRICT+focus on‘transfer of good practices and policies improved'65 Main conclusions and recommendations:
by moving towards concrete transfer of identified good practices (already available in the ERIK database) into mainstream Structural Funds programmes in regions wishing to improve policies.
The students are selected annually via a database of at least 350 students from all over the world (mostly Swedes).
while several programmes and measures exist in many forms across Europe to assist SMES with the transition towards a digital economy,
Overtext Web Module V3.0 Alpha
Copyright Semantic-Knowledge, 1994-2011