93 years of economic insights for Indiana

The IBR is a publication of the Indiana Business Research Center at IU's Kelley School of Business

FDI Announcements: A Potential Signal of the Benefits of Cluster Development

Research Associate, Indiana Business Research Center, Indiana University Kelley School of Business

Research Associate, Indiana Business Research Center, Indiana University Kelley School of Business

Director of Economic Analysis, Indiana Business Research Center, Indiana University Kelley School of Business

Clusters—which are a co-location of firms that may share supply chains, modes of transportation or similar talent requirements—provide a network of economic relationships in a region that can create a competitive advantage for related firms (like software firms congregating in Silicon Valley). Industry clusters, it is said, facilitate the exchange of supplies, personnel and information between related firms in a region.1 Cluster growth, therefore, may be important for the economic well-being of the region as a whole.

These clusters of industries that are growing in a region are aptly named “growth clusters.2 In earlier issues of the Indiana Business Review, we have also written about the role of cluster-based development.3

Clusters grow because those industries in a particular region have a competitive advantage. Differing forces may be at play in that growth. They can grow “metabolically,” that is, expand using the resources, labor and know-how in the region, as well as technology from outside the region—combined with increasing demand for the cluster’s goods and services outside the region. Clusters can also grow “magnetically,” that is, a region can attract firms to take advantage of that region’s competitive advantage in resources, supply networks or human talent. An example of magnetic growth is attracting foreign direct investment (FDI).

Michael Porter suggests that clusters may attract FDI by providing easy access to resources, technologies and markets, though other scholars are quick to point out that clusters and FDI can be interdependent phenomena. Clusters may also have an influence on the foreign companies that are doing the investing in the region.4 The impact of cluster-related FDI on the wider economy also renders the relationship between clusters and investment particularly important. Does FDI stimulate clustering activity and generate positive spillover effects into the wider economy? Some researchers have found that productivity spillovers from FDI actually occur only in pre-existing clusters, suggesting that the relationship between clusters and FDI is complex and worth exploring in greater detail.5

In this article, we will explore the role that growth clusters may have in attracting FDI, i.e., magnetic cluster growth, in Indiana. We used industry cluster definitions from the U.S. Cluster Mapping Project and employment by industry data from QCEW-complete employment estimates, which the Indiana Business Research Center (IBRC) provides on the Hoosiers by the Numbers website. A proprietary data set, fDiMarkets, is the source for announced or “intended” FDI flows. All of these data are at the county level.

In contrast to other FDI data sources, fDiMarkets data are comprised of press releases from firms and economic development agencies that announce an intended greenfield investment or an expansion of existing facilities. Merger and acquisitions (M&A) are not included. For the purposes of our inquiry, intended greenfield investments signal a company’s intention to locate a new facility (or expand an older one), with the emphasis on new. Our contention is that greenfield investment intentions are a stronger signal than M&A activity for the sake of measuring a region’s magnetism. M&A flows of FDI change the owner on the masthead, but the effects are not as apparent.

There are contrasting views on this. A foreign firm buying an existing operation may be motivated by the firm wanting to get a piece of a cluster’s competitive benefits in a region. The Brookings Institution released a report in the summer of 2014 noting the advantages of M&As in terms of job growth, beneficial spillover effects and regional vitality, and provided some anecdotal evidence.6 The report also cites several articles related to FDI and productivity growth. That said, according to one article, finding robust empirical evidence to support positive spillovers is more difficult than finding theoretical reasons spillovers may occur.7 Perhaps the better case for why greenfield FDI is a better magnetism indicator than M&A investment is that few economic development officials devote a majority of their time soliciting foreign firms to buy out local companies.

Results

For the first of several empirical analyses, we used a statistical procedure that estimates the probability that the presence of a growth cluster attracts greenfield FDI.8 The results of the statistical analysis do not point to a strong relationship between the presence of clusters and attracting FDI. The first statistical model shows that the presence of a traded growth cluster (having an employment concentration greater than the national average) alone increases the likelihood of FDI flows into the county by about 39 percent. That said, the presence of a growth cluster does not appear to explain much of the variation of FDI flows. The decision of making an FDI investment or not that can be explained by the simple model—pseudo R2 for statistical buffs—is only 0.005. In other words, only 0.5 percent of the variation in the FDI decision can be explained by the presence of a traded growth cluster in the region. (The full table with results from all models can be found in the appendix at the end of this article.)

The simple statistical model addresses only a simple, binary, “on-off” world. That is, either there is a growth cluster or not and either there was FDI or not. If scale, or the size of the growth cluster, is taken into account, more of the variation in the FDI decision can be explained. The second statistical model shows that it is not so much the presence of a cluster that matters, but rather that the scale of the growth cluster (measured by employment) explains more of the variation in the FDI decision. The model indicates that a 1 percent increase in the average employment share of a given growth cluster would nearly double the likelihood of attracting foreign investment.9

The first two models were generic, industrially speaking. In other words, all growth clusters and the industries that comprise these growth clusters were treated the same. All of the cluster employment and all of the FDI were aggregated together. In two additional models, we examined the role that the specific industry and cluster may play in attracting FDI. We looked at several clusters of particularly high employment concentration in the state: automotive, biopharmaceuticals, upstream metals, recreational vehicles (RV) and medical devices.10 It turns out that the RV cluster (centered around Elkhart) had no RV-related FDI (see Table 1).

Table 1: Cluster Presence and FDI Inflows by Industry, 2007 to 2013

Cluster Number of Counties Not Attracting FDI Number of Counties Attracting FDI Total Number of Indiana Counties
Automotive 9 28 37
Local Health Services 22 17 39
Upstream Metals 15 16 31
Medical Devices 6 6 12
Biopharmaceuticals 7 2 9
Recreational Vehicles (RVs) 14 0 14

Source: IBRC, using fDi Markets data

This may be attributed to the gutting of the RV industry during the Great Recession, as our data covers the period from 2007-2013. Perhaps during these years, RV manufacturing was not considered an attractive industry for foreign direct investment. The RV industry, as a result, did not warrant further statistical analysis. Additionally, in order to focus on traded clusters (i.e., those that sell to other regions), we omitted the local health services cluster from the analysis as well.

Figure 1 shows the relationship between the industry growth cluster’s average location quotient and the probability of attracting FDI for the four remaining key industries from 2007 to 2013. Bubble size indicates the average level of employment in that growth cluster over this time period. Though the medical devices growth cluster has the largest LQ (as indicated by its position to the far right on the graph), the biopharmaceutical industry has the greatest probability of attracting FDI, as 100 percent of the biopharma growth clusters attracted FDI; the automotive industry comes next, with a 53 percent probability of attracting FDI, followed by medical devices (50 percent) and upstream metals (43 percent).

Figure 1: Predicted Probability of FDI in Four Growth Clusters by Average Location Quotient, 2007 to 2013

graph

Note: Bubble size indicates average cluster employment.
Source: Indiana Business Research Center

The third statistical model was motivated by the hypothesis that FDI attraction may be affected by the industry, that is, the specific industry’s likelihood of attracting FDI. This model shows that there are clear differences by industry. Growth clusters in the automotive sector attracted FDI to an Indiana region (such as Honda opening a new factory in Indiana), while the upstream metal sector also attracted FDI. The odds of these industries attracting FDI are well over 100 percent in our model, with a growth cluster in the automotive industry being 431 percent more likely to attract FDI than a growth cluster in any other industry. A growth cluster in the upstream metal industry is 177 percent more likely to attract FDI than a growth cluster in any other industry (see Table 2). The biopharmaceuticals and medical devices industries did not have any statistically confirmed effects on the odds of attracting FDI.

Table 2: Net Effects of Cluster on the Likelihood of Attracting FDI, 2007 to 2013

Industry Sector Likelihood of Attracting FDI
Automotive 431%
Biopharmaceuticals No Effect
Medical Devices No Effect
Upstream Metals 177%

Source: Indiana Business Research Center

Our fourth model of selected industrial sectors looks at the effect of growth cluster strength, using the interaction between a cluster’s location quotient—that is, the relative concentration of employment in a cluster—and the presence of a growth cluster in the region. The model suggests that the presence of upstream metal manufacturing will indeed attract FDI to a region, as will the presence of a biopharmaceutical growth cluster. The latter result fits with the fact that two regions with growth clusters in biopharma attracted all of the biopharma FDI. No weak biopharma cluster regions attracted FDI, as Table 3 shows. Table 3 also signals why the industry results in Model 4 may not be particularly strong for industries other than biopharma. Namely, weak cluster regions also attracted FDI. That said, as column five in Table 3 indicates, the share of the dollar value of those FDI flows into weak cluster regions were not high. (One minus the percentage in column five equals the percentage of investment dollars flowing into weak cluster regions.)

Table 3: FDI Attracted by Growth Cluster Regions for Selected Industry Clusters, 2007 to 2013

Cluster Total Number of Regions Growth Clusters Weak Clusters
Total Number of Regions Number of Regions Attracting FDI Dollar Share of FDI (Percent) Number of Regions Attracting FDI
Automotive 37 34 26 96% 2
Biopharmaceuticals 9 6 2 100% 0
Medical Devices 12 8 4 85% 2
Upstream Metals 31 28 14 95% 2

Note: Weak clusters are those with employment concentration less than the United States. Growth clusters are those with high employment concentration (high LQs).
Source: Indiana Business Research Center

Conclusion

For Indiana, the evidence suggests that the presence of growth clusters plays, at best, a marginal role in attracting greenfield FDI. Higher average employment concentrations are a better predictor of FDI flows. The degree to which clusters attract FDI—growth via magnetism—appears to be limited in a generic industry sense, but appears to be sensitive to the type of industries that make up the cluster, e.g., the automotive industry. It is important to consider not just the presence of growing clusters that may attract additional outside investment, but the cluster’s size and industry composition that may influence magnetic growth. The analysis was limited to Indiana, and as a result can’t be generalized to the nation as a whole. However, these findings provide possible questions for additional research on cluster growth on a national scale.

Appendix

Model 1 shows the mean difference of the odds ratio between growth and non-growth (weak) clusters in terms of attracting FDI. Models 2 through 4 show that, after controlling for the average employment and specific industry clusters (and whether they are growth clusters), the presence of a growth cluster alone loses explanatory power in attracting FDI. The highly significant and large positive (negative) effect from the presence of biopharmaceutical growth clusters (and the sector in general) in Model 4 reflects the fact that all FDI in the biopharmaceutical sector went to its growth clusters. In other words, being a non-growth (or weak) biopharmaceutical cluster does not attract FDI at all.

Table 4: Effects of Growth Clusters on the Likelihood of Attracting FDI

  Model 1 Model 2 Model 3 Model 4
Location Quotient (LQ) 0.328** 0.011 -0.166 -0.145
  (0.15) (0.16) (0.17) (0.17)
Shift-Share Industry 0.123 0.226 0.239 0.246
  (0.16) (0.18) (0.18) (0.18)
Average Employment   0.681*** 0.658*** 0.657***
    (0.07) (0.07) (0.07)
Automotive Industry     1.669*** 1.685
      (0.40) (1.26)
Automotive × LQ       -0.026
        (1.33)
Biopharmaceutical Industry     -0.753 -12.747***
      (0.78) (0.62)
Biopharmaceutical × LQ       12.256***
        (1.08)
Medical Devices Industry     0.716 1.176
      (0.62) (0.99)
Medical Devices × LQ       -0.736
        (1.24)
Upstream Metal Industry     1.020** 1.892*
      (0.38) (1.00)
Upstream Metal × LQ       -0.966
        (1.08)
Constant -1.016*** -4.965*** -4.858*** -4.868***
  (0.12) (0.43) (0.44) (0.44)
Observations 863 863 863 863
Pseudo-R2 0.005 0.118 0.142 0.144

Note: The dependent variable is a binary indicator for FDI. The independent variable of interest is also a binary indicator for being a growth cluster or not. The model is estimated using logit models for traded industrial clusters only. The selected industrial clusters in Model 3 include automotive, biopharmaceuticals, upstream metal and medical devices. Local health services and RV clusters are excluded from the model because the former is not traded and the latter has no FDI. The interaction term between the biopharmaceutical sector and the growth cluster is also excluded in Model 4 because clusters that have FDI are all growth clusters within the sector. Standard errors are in parentheses. Significance level: * 10%, ** 5% and *** 1%.
Source: Indiana Business Research Center

Notes

  1. M. Porter, “Clusters and the New Economics of Competition,” Harvard Business Review, November-December 1998, https://hbr.org/1998/11/clusters-and-the-new-economics-of-competition.
  2. It is important to differentiate the notion of cluster definitions, which are groupings of industries that have been set forth by Porter and the U.S. Cluster Mapping Project, and the presence of a cluster in a region, which is based upon the relative concentration of employment for a particular grouping, or cluster of, industries.
  3. T. Slaper and G. Ortuzar, “Industry Clusters and Economic Development,” Indiana Business Review, Spring 2015, www.ibrc.indiana.edu/ibr/2015/spring/article2.html.
  4. P. Gugler and S. Brunner, “FDI Effects on National Competitiveness: A Cluster Approach,” International Advances in Economic Research 13 (2007): 268-284.
  5. L. De Propris and N. Driffield, “The Importance of Clusters for Spillovers from Foreign Direct Investment and Technology Sourcing,” Cambridge Journal of Economics 30, no. 2 (2006), http://cje.oxfordjournals.org/content/30/2/277.
  6. D. Saha, K. Fikri and N. Marchio, “FDI in U.S. Metro Areas: The Geography of Jobs in Foreign-Owned Establishments,” June 2014, http://www.brookings.edu/~/media/research/files/reports/2014/06/20-fdi-us-metro-areas/metrofdi.pdf.
  7. H. Gorg and D. Greenaway, “Much Ado about Nothing? Do Domestic Firms Really Benefit from Foreign Direct Investment?” (working paper, Institute for the Study of Labor, November 2003), http://papers.ssrn.com/sol3/papers.cfm?abstract_id=475044.
  8. We performed a logit regression to determine the probability of clusters attracting FDI. The logit modeling method produces a result that can be interpreted as the likelihood of one condition leading to another, making it suitable for modeling the effect that clusters have on attracting foreign investment. The presence of a traded cluster—i.e., those industries that generally make for and sell to consumers outside the region—is determined by a location quotient (LQ). An LQ measures the concentration of employment by comparing the percent of employment in a given region in Indiana to national averages. Therefore, the presence of a cluster in an Indiana county is indicated by a higher percentage of workers employed in that industry than is typical across the U.S. as a whole. FDI is measured as a binary variable: counties that have received foreign direct investment are contrasted with those that have not received any, regardless of the differences in actual dollar amounts invested in each county.
  9. Controlling for the average employment for each cluster from 2007 to 2013, the impact of the cluster presence variable is reduced dramatically and loses its statistical significance in Model 2. In turn, the effect of cluster employment on FDI is highly significant.
  10. The upstream metal cluster includes pipe, tube and rolled steel manufacturing, forging and secondary smelting industries. The official term for the recreational vehicle cluster is “trailers, motor homes and appliances,” and is overwhelmingly dominated by RV manufacturing in Indiana (hence the lay terminology).