Discussion of Identifying Neighborhood Effects Among Firms: - - PowerPoint PPT Presentation

Discussion of “Identifying Neighborhood Effects Among Firms: Evidence from Location Lotteries of the Tokyop Tsukiji Fish Market ” by Kentaro Nakajima and Kensuke Teshima

Jessie Handbury, Wharton

CJEB Japan Economic Seminar, February 16, 2018

2

Store Placement

• Top three in mall sales:
• 1. Women’s clothing (16.6%)
• 2. Shoes (9.9%)
• 3. Food (9.8%)
• Where would you place

these stores in this mall?

Summary

• Question: Do stores agglomerate for the sake of demand spillovers?

– Old theory behind the spatial organization of retail – Why? Search costs (macro), trip chaining (micro)

• Identification Challenges:

– Selection of firms into high-demand locations Solution: relocation lotteries in wholesale fish markets – Production externalities Solution: differentiated effects from front-facing vs. back-facing neighbors

• Clean identification of firm-level benefits of demand spillovers

– Jardim (2016) identifies intra-firm spillovers separately from location attributes – Relihan (2017) measures demand spillovers directly with shopping data

Results

• Small, specialized stores benefit from demand spillovers

– One additional fish specialty neighbor increases the probability of expansion by 25% from the baseline (9%) – One standard deviation increase in diversity increases probability of expansion by 45%, equivalent to effect of being a (highly visible) corner store

• Zero evidence of supply spillovers (even within-group)
• No positive demand spillovers for multi-shop retailers

• where:

– Δ yigr = change in the number of stores applied for between 1990 and 1995

• exitigr if firm exits between 1990 and 1995 lotteries
• Δ Shopsigr difference in number of shops applied for, conditional on survival
• I[Δ Shopsigr >0] dummy for whether the number of shops increased between

the 1990 and 1995 applications

– Diversityr = number of trade groups represented in the neighborhood or HHI within neighborhood as the result of 1990 lottery

• Dynamics:

– Shop’s position between 1990 and 1995 determines its expected sales in 1995 lottery position. – Either buyers find them and stick with them or they use the increased revenues in 1990 position to invest in technology.

• where:

– Δ yigr = change in the number of stores applied for between 1990 and 1995

• exitigr if firm exits between 1990 and 1995 lotteries
• Δ Shopsigr difference in number of shops applied for, conditional on survival
• I[Δ Shopsigr >0] dummy for whether the number of shops increased between

the 1990 and 1995 applications

– Diversityr = number of trade groups represented in the neighborhood or HHI within neighborhood as the result of 1990 lottery

• Space:

– “Neighborhood” = area separated by larger corridors – Implicit assumption: effects operate within but not across regions – Implication: corner (“border”) stores are less valuable, not more – Why not use a broader neighborhood definition to account for wider effects?

• Trade-off between clean identification and generalizability

– Finding for corroborating evidence for null results (lack of supply-side spillovers) in the broader Japanese retail market might be tough. – Start by looking at another fish market with different allocation mechanism? – Instead motivate with spatial distribution of Japanese retail in Census data?

• Do the results suggest an efficient spatial organization for the market?

– e.g., tuna stores = anchor for sushi fish wholesalers

• How much do stores value demand externalities?

– Can pre-trading secondary market be used to measure the role of consumption externalities in firm agglomeration? – Is there sorting into group applications (self-selected horizontal neighbors)? – Do other fish markets auction spaces?

Economic Significance

8

Source: WSJ, November 24, 2015