Replication Plan

Each student team will conduct an exact replication and at least one of the below proposed constructive replications. All teams will jointly analyze and publish overall findings.

Original Study

Kim, M. J., Shaver, J. M., & Funk, R. J. (2022). From mass to motion: Conceptualizing and measuring the dynamics of industry clusters. Strategic Management Journal, 43(4), 822-846. https://doi.org/10.1002/smj.3354. 

Original Study Research Question and Findings

Industry clusters are geographic concentrations of firms from a specific industry. Kim, Shaver, and Funk (2022) introduced and validated a measure of Cluster motion, which captures the change in the level of concentration of firms in a specific region over time. They validated this measure for two industries, the computer and the semiconductor industry using US Census data. Descriptive results suggested that regions rarely followed stylized descriptions of cluster life cycles, and instead, the authors identified five different concentration change patterns over time. 

Data

Industry categorization and geographic location information for all business establishments in the semiconductor and the computer industries based on publicly available U.S. Census data for 1974–2016.

Design and Analyses

The three-step process to estimate changes in regional industry concentration levels (Cmotion):

1. Determine annual industry concentration for each U.S. Metropolitan Statistical Areas (MSAs) by first using Monte Carlo simulations to randomly distribute the total number of business establishments in the US randomly across all Metropolitan Statistical Areas (MSAs). Then, use the observed number of firms in each MSA-year to calculate z-scores, which quantify the extent to which the degree of concentration in a focal MSA-year differs from what would be expected by chance.

2. Structural break analysis (Bai & Perron, 1998, 2003) identifies changes concentration trends for each MSA over time.

3. Regression analysis estimated linear concentration trends with coefficient estimates capturing the direction and magnitude of change (Cmotion). Figure 1 shows two examples of dynamic concentration patterns identified by Kim et al. (2022).

Reproduction

• Using author-posted data and code.

Exact / Literal Replication

• Using US Census Data for the same time period (1974-2016).

Constructive Methods-Focused Replications

• Updated time frame using US Census Data (1974-2020)

• Alternative measures of economic activity (employment instead of establishments)

• Alternative MSA Weights (land area instead of resident population)

• Alternative cluster definitions (related industries instead of SIC codes)

Constructive Generalizability-Focused Replications

• Validate introduced Cmotion measure for alternative industries with different innovation focus and scale economies.

• The original study reported that clusters rarely follow simple concentration life cycle models. Validate the following five alternative patterns of cluster dynamics reported in the original study for alternative industries and the extended time frame.

• Identified patterns of cluster dynamics (see also Figure 2):

  1. concentration => dissipation

  2. long period of concentrations => short period of dissipation

  3. concentration => dissipation => concentration

  4. increasing concentration => stability

  5. stable period/cluster emergence => increasing concentration

Constructive Theory Extension Study

• Regional additive effects across industries.

• If regional outcomes reflect cluster dynamics at the industry level or reflect some “superadditive” effect across industries within a region. Such analysis might be a path towards distinguishing the relative importance of Marshallian (i.e., specialization) vis-à-vis Jacobian (i.e., urban) externalities.