For a useful discussion of spillovers and possible transmission mechanisms, see Markusen’s (1998) review of multinationals. See also Caves (1974), Globerman (1979) and Baldwin and Hanel (2003).

For an overview of micro-economic research on productivity dynamics, see Bartelsman and Doms (2000) that includes a more comprehensive discussion of many of these points. For a summary of the issues involved in measuring the impact of turnover on productivity growth, see Baldwin and Gu (2006).

Earlier studies (Baldwin 1995) find similar orders of magnitude.

The contribution of firm turnover to aggregate productivity growth is even higher in the retail sector (Baldwin and Gu 2008).

This characterization of spillovers as an involuntary transmission of benefits—productive knowledge not fully appropriated by the firm generating the externality—is found in Markusen (1998).

Foreign firms are not the only source of technological spillovers. Griffith, Redding and Simpson (2002, p. 24) also report that “high productivity foreign-owned establishments are found to make an important contribution to productivity growth by pushing out the technological frontier, but high productivity UK-owned firms, to the extent that they are also technological leaders in some industries, are also important sources of technology transfer.” Baldwin and Gu (2005, p. 6) present similar findings, arguing that the “foreign ownership advantage is found to be a multinational advantage;” the authors note that Canadian multinationals operating in Canadian manufacturing industries exhibit technology, wage and productivity profiles on par with foreign multinationals.

We should note that testing for spillover effects was not the primary focus of Baldwin, as his analysis focused on estimating the contribution of plant turnover to aggregate productivity growth (see 1995, chap. 9).

Some analysis of transmission mechanisms is found in Baldwin and Gu (2005). Using a linked dataset that combines longitudinal data on plant performance with data collected from a technology survey, the authors test for the presence of spillovers by examining whether cross-industry variation in the share of output under foreign control influences (a) the level of competition faced by domestic plants, and (b) the extent to which these plants adopt advanced technologies. In both cases, the impact of foreign control is found to be positive and significant, suggesting that competition and technology use are areas in which domestic companies benefit from the co-existence of foreign-controlled plants.

The version of the Longitudinal Manufacturing Research File (LMRF) used here covers the years from 1973 to 1999 inclusive with a constant Standard Industrial Classification industry code. Many of the studies that have utilized the LMRF are published in Statistics Canada’s Economic Analysis Research Paper Series.

These spatial data were developed at Statistics Canada to support a series of research projects that have examined the link between urban agglomeration and productivity growth (e.g., Baldwin et al. 2007; Baldwin, Brown and Rigby 2008).

See CANSIM table 383-0022. These estimates of annualized growth are based on value added.

For each of our analysis samples, the incumbent average is calculated as the weighted average level of labour productivity for all continuing plants operating during the observation period. This incumbent population is effectively the union of market-share gainers and market-share decliners as described herein, as these tabulations exclude the impact of entrants and exits.

Productivity growth for plant i is calculated as the compound annual growth rate over the analysis period. Labour productivity for plant i is the ratio of real manufacturing production to plant employment. We use production per worker rather than value added per worker in order to reduce the number of aberrant observations in the microdata due to reporting errors. Previous studies have produced quite similar results at a more aggregated industry level using both formulations and our experiments produced the same result.

Note that these descriptive tabulations do not take differences in industry membership into account.

Again, we reserve any rigorous consideration of industry effects for the formal econometric model.

As noted in Section 3, our measure of productivity growth is the plant’s compound annual rate of productivity growth over the observation period.

This ranking was based on all plants operating in the start year of the analysis period—both continuers and non-continuers.

The use of productivity gaps to operationalize the distance from the frontier is also employed by Griffith, Redding and Simpson (2002), who employ a measure of total factor productivity.

We are tacitly assuming that the source of these shared effects is exogenous to both market-share gainers and market-share decliners. A full consideration of these shared effects—beyond our scope here—would necessarily ask whether they are truly exogenous. If some industries experience more rapid productivity growth than others, and this higher productivity emanates directly from the actions of market-share gainers, then these industry conditions do ostensibly represent a type of spillover—as gainers are creating the conditions that enable struggling firms to prosper. These distinctions underscore the challenges of conceptualizing spillovers within dynamic systems of competitive reallocation.

Geographic areas were defined as census metropolitan area/census area/census divisions.

A detailed discussion of the edit rules that were applied to our data samples, along with descriptive statistics, is found in Appendix A.

We have ignored the impact of the frontier because it has a large negative effect when it is significant, thereby suggesting that it subtracts from rather than contributes to the centripetal effect of the externality.

See Baldwin and Rafiquzzaman (1994) for a discussion of the methodology used to create these groupings.