| Abstract: |
Industrial diversity is widely held to buffer urban economies against economic
shocks, yet evidence remains mixed and the operative mechanisms poorly
understood. Using establishment-count data for 131 wards across nine Japanese
designated cities and Tokyo (2016--2021), we apply Bayesian spatial error
models to examine how three diversity dimensions---within-ward breadth
($\alpha$-diversity), between-ward compositional heterogeneity
($\beta$-diversity), and related/unrelated variety---independently predict
ward-level resilience during the COVID-19 shock. Related diversity is robustly
protective while unrelated diversity is harmful---a sign reversal relative to
long-run growth studies that we attribute to the sector-discriminating
character of the pandemic: cross-sector portfolio breadth provides no hedging
mechanism when proximity-dependent industries collapse simultaneously. Shannon
entropy's positive association with resilience masks this opposition between
related and unrelated variety. $\alpha$-diversity and sectoral concentration
(HHI) exhibit mutual suppression, implying that industrial breadth confers
resilience specifically when organised around an anchor sector, while
$\beta$-diversity proves to reflect spatial clustering of the information and
communications technology sector rather than general compositional
heterogeneity. These results imply that the policy-relevant target is
strategic \emph{related} diversification---broadening within-sector-family
portfolios---rather than maximising generic industrial breadth, and that the
operative diversity mechanism depends critically on the type of shock a city
faces. |