6. Innovation In an Integrated Framework: A Europe-United States Comparative Analysis

“A patent is a member of the triadic patent families if and only if it has been applied for and filed at the European Patent Office (EPO), at the Japanese Patent Office (JPO) and if it has been granted by the US Patent and Trademark Office (USPTO)” (Eurostat 2006: 6). Patent families are supposed to improve international comparability by suppressing the home advantage.

€ 8,049.5 in the EU-25 and € 8,422.6 in the Euro Zone vs. € 20,487 in the US, measured in PPS, based on full-time equivalents.

When the ranking is extended to the top 100 universities we find that 57 are in the USA and 35 in the EU (of which 11 in the UK). The ranking of the top 500 universities in the world is based upon a variety of performance indicators (see http://​ed.​sjtu.​edu.​cn/​rank/​2005/​ARWU%20​2005.​pdf for further details).

Though the effects of the 1982 patenting system reform are debated (see Jaffe and Lerner 2004).

Duranton and Puga (2003) use as an example a model “in which agglomeration facilitates the matching between firms and inputs. These inputs may be labelled workers, intermediates, or ideas. Depending on the label chosen, a matching model of urban agglomeration economies could be presented as a formalisation of either one of Marshall’s three basic sources of agglomeration economies even though it only captures a single mechanism” (p. 2).

Zimmermann (2005) points out that the EU shows “a split labour market that is characterized by high levels of unemployment for low-skilled people and a simultaneous shortage of skilled workers. This lack of flexible high-skilled workers and the aging process has created the image of an immobile labour force and the eurosclerosis phenomenon (thus preventing) the best allocation of resources and hence economic efficiency” (p. 448).

For a review of the theoretical and empirical works based on this approach and for a discussion of its limitations see Wieser 2005.

The majority of patents issued by the USPTO are utility (i.e., invention) patents. Other types of patents and patent documents issued by USPTO, but not included in this report, are plant patents, design patents, statutory invention registration documents, and defensive publications. While in 1999 the number of utility patents granted reached 153,493, just 14,732 design patents, 448 reissue, and 421 plant patents were awarded. Our data do not include these other categories.

The USPTO provides data at the sub-state level on utility patents granted from 1990 to 1999 with a first-named inventor who resided in the United States. For the EU, instead, patents are organized by EUROSTAT according to the application years rather than the grant years. However, the US patent data at the national level show that the numbers of patent applications and patents granted are highly correlated over time (0.94 for the period 1989–2002) and across geographical units (0.98 for 1990).

As the time distance-matrix is calculated either at the NUTS1 or at the NUTS2 level, in order to make it coherent with our data which combine different NUTS levels we relied on the NUTS distance matrix using the NUTS 2 regions with the highest population density in order to represent the corresponding NUTS1 level for Belgium, Germany, and the UK.

The distance matrix does not take into account the impact of railway and/or air connections on the average trip length. Only road travel-time is available for the EU regions.

Data on distances between MSAs are calculated on the assumption that a 1 degree difference in latitude is constant regardless of the latitude being examined. This assumption is not problematic for smaller countries, but for a large country like the US, it may result in an underestimation of the distance between Southern cities and an overestimation of that between Northern cities.

The 1990 census classification was developed from the 1987 Standard Industrial Classification (SIC) Manual published by the Office of Management and Budget Executive Office of the President.

The first category includes people whose highest level of schooling is an associate degree (for example: AA, AS) or some college credit, but no degree. The second group includes those whose highest level of schooling is a bachelor’s degree (for example: BA, AB, BS), a master’s degree (for example: MA, MS, MEng, MEd, MSW, MBA); or a professional degree (for example: MD, DDS, DVM, LLB, JD) (US Census Bureau).

Unemployment is “hidden” in the fabric of very small farm holdings in many EU peripheral areas and in many Southern states of the US (Demissie 1990; Caselli and Coleman 2001). In both these contexts agricultural workers show low levels of formal education, scarce mobility, and tend to be aged.

We are aware of the potential endogeneity arising from the introduction of the social filter variable into the Knowledge Production Function. An effective strategy in order to address this issue would imply the use of several time lags of the social filter variables as instruments in an instrumental variables framework. However, due to the constraints on data availability discussed before, we are forced to limit ourselves to considering the value of this indicator at the beginning of the period of analysis, while assessing the patent growth rate over subsequent years.

Migration data are provided by Eurostat in the “Migration Statistics” collection. However there are no data for Spain and Greece. Consequently, in order to obtain a consistent measure across the various countries included in the analysis, we calculate this variable from demographic statistics. “Data on net migration can be retrieved as the population change plus deaths minus births. The net migration data retrieved in this way also includes external migration” (Puhani 2001: 9). The net migration was standardised by the average population, obtaining the net migration rate. Consequently, while for the EU it is impossible to distinguish between national, intra-EU, and extra-EU migration flows, for the US domestic in-migration and out-migration data consist of moves where both the origins and destinations are within the United States.

Different estimation techniques have been considered in order to minimize potential bias due to omitted variables (panel data) and/or modifiable aerial unit problem (e.g., Hierarchical Linear Models, HLM). However, severe constraints in terms of both the spatial scale and the time-series dimension of the existing regional data prevent us from implementing these alternative methodologies. Further research in this direction remains in our agenda, given the continuous progress in the availability of regional statistics.

The MSA/CMSA list is based on Metropolitan Areas and Components, 1993, with FIPS Codes, published by the Office of Management and Budget (1993).

Standard & Poor’s Compustat North America is a database of financial, statistical, and market information covering publicly traded companies in the U.S. and Canada. It provides more than 340 annual and 120 quarterly income statements, balance sheets, flows of funds, and supplemental data items on more than 10,000 active and 9,700 inactive companies.

The concept of FURs has been defined as a means to minimize the bias introduced by commuting patterns. A FUR includes a core city, where employment is concentrated, and its hinterland, from which people commute to the center. For a detailed analysis of this concept see Cheshire and Hay (1989).

The 145 MSAs for which R&D data are available account for 89,9% of the GDP generated in all 266 MSAs and show an average of 225.19 patents per million inhabitant against 176.83 for the whole sample.

Beeson et al. (2001) discuss the “sample selection bias” introduced when choosing cities as unit of analysis rather then county-level data: only places that experienced successful growth in the past are considered in this way. The use of standard metropolitan statistical areas minimises this first bias. However, in order to keep the bias at a minimum, we not only report the results for the most innovative subsample of MSAs, but also for all MSAs in the continental US.

When assessing this phenomenon it must, however, be borne in mind that the unit of analysis in the case of the EU are NUTS regions i.e., territorial units for the production of regional statistics for the European Union whose definition mainly serves administrative purposes. As a consequence, NUTS regions might not always approximate the functional borders of the regional economy. Conversely, US MSAs are closer to the concept of ‘functional urban regions’ (Cheshire and Hay, 1989) and likely to be more “self-contained” in terms of economic interactions. Consequently, part of the difference between the empirical evidence recorded in the two cases may be due to the different nature of the spatial unit of analysis. However, since we rely on inverse linear distance (and not on contiguity) for the specification of our spillovers variable, the impact of heterogeneous spatial units in Europe and the US should be minimized supporting our results that are, in any case, largely in line with the existing literature. As a robustness check we have re-estimated our empirical model for the US, using state-level data (contiguous) with very similar results for the spillover variables. These results are available from the authors on request.

This is consistent with the notion that because mobility is higher in the US, innovation systems have more local matching and learning and hence are more “local” than in Europe, where long distance communication is necessary in order to match relatively immobile agents.

The impact of the sectoral structure upon regional innovative performance cannot be limited to the overall degree of specialisation. On the contrary, it would be necessary to fully control for the specific regional patterns of specialisation: given the aggregate degree of regional specialisation, the true differential factor could stem from a region being specialised in high tech R&D-intensive vs traditional sectors (Smith 2007). Further investigation, in an EU-US comparative perspective, of the sectoral level territorial processes remains in our agenda for future research.