Acquisitions, Management and Efficiency in Rwanda’s Coffee Industry

Discussion by J. Anthony Cookson

Overview

• Misallocation & inefficiency are important obstacles for firms in low-income countries (Hsieh and Klenow 2009, Bloom et al 2012).

​

• Can acquisitions help improve efficiency?
• Braguinsky et al (2015) and Maksimovic and Phillips (2001) suggest this could be an important channel.
• But evidence is scant, partly because acquisitions are rare in a developing context.

​

• If acquisitions improve efficiency, what kinds of acquisitions are most valuable?
• What channels enable efficient acquisitions? Access to capital, technology, management practices… etc.

This paper

• Studies a recent wave of consolidation in the coffee mill industry.
• Finding – acquisitions improve efficiency of plants.

​

• Not all plants do as well.
• Foreign-acquired plants are systematically more efficient.
• Half of the gains due to management practices like pay and hiring better managers.

​

• Other half is most consistent with foreign firms facing less resistance in implementing changes.

​

Take and Discussion

Take

• Interesting question.
• Understanding how weak institutions relate to management practices is an important intersection of ideas
• Ambitious data collection, part of an impressive multi-paper effort.

​

Discussion

• Puzzling about the resistance result.
• Diff-in-diff? Counterfactuals? Selection vs. Treatment
• Interpretations of channels

Comment 1: Role of informal institutions

Outsiders face less resistance?

• The finding that domestic firms face less resistance in implementing changes is surprising.

​

• From an institutional viewpoint, you would expect local/domestic firms to have more sway
• E.g., Rwanda is a country where informal norms would seem to be strong.

​

Comment #1: Suggestion

• What is the logic for why foreign firms face less resistance in implementing change?
• Unpack this result much further – not just follow-on tests.
• Much of this is exposition, but more institutional detail too.

​

• Do foreign firms provide more clarity in decision making?
• If a domestic firm and a foreign firm propose the same change, are there informal mechanisms for employees to push back against domestic firms?

​

​

​

​

Comment 2: Diff-in-diff

Empirical strategy�Difference-in-difference around acquisitions

Diff-in-diff is at its best when the event is outside of the control of the firm.

• That’s not the case here
• Robustness to convince that selection of targets isn’t just picking mills that would perform well in the absence of the acquisition.
• E.g., showing parallel trends, carefully picking counterfactual non-acquired targets.

​

More on the counterfactuals

• Acquirer survey allows to restrict comparisons
• Foregone: other acquisition targets reported by the same acquirer.
• Possibly foregone: mills that existed at the time of the acquisition and that the acquirer would have considered as alternative targets.

​

• The problem with either approach is:
• The acquirer forewent these potential targets because of the selection effect.
• Bigger estimates in these specifications suggest an important selection effect, rather than assuaging concerns.
• In fact, in Akkus et al (2016; 2021 Mgmt Science), we use variation like this explicitly to quantify selection.

​

Suggestion for Comment #2

• I would drop diff-in-diff terminology entirely.
• As in Braguinsky et al (2015), the most natural interpretation is that the acquisition improved the performance.
• I believe the central findings, but the diff-in-diff terminology does you no favors.

​

• It would be simpler to say:
1. After the acquisition, acquired mills performed better than non-acquired.
2. Before the acquisition, they performed worse, and they weren’t improving.

​

• These are the results, and they’re the reason you’d believe that the acquisition led to the performance improvement.
• After “leading the horse to the water,” you could say that this satisfies diff-in-diff, but the narrative wouldn’t raise such high expectations.

Where did I get this idea?

• From Braguinsky et al (2015) … the first time they mention difference-in-differences is page 14 in footnote 20 (!).
• They also don’t even say they’re doing diff-in-diff. They’re using a framework “similar to difference-in-difference” estimation (page 16).

Comment 3: Channels

Evidence related to three channels

1. Foreign firms hire better managers & pay them more.
• Table V and VI

​

• Foreign firms face less resistance.
• Table VIII

​

• Foreign firms do not confer financing nor technological advantages.
• Appendix evidence

Important diffs between foreign and domestic

• Foreign firms are bigger, interacting with 21 mills vs 4 mills on average.

​

• Are there advantages to being big?
• In Rwanda, do you get sway with government officials?

​

• Coordination across mills / assets?

​

• Internal capital markets?

Evidence on the internal finance channel

Placing a lot of weight on a non-significant difference

39 observations, and only 8 foreign exporters.

​

• Standard errors ~20pp in cols 1 and 2.
• Lower bound on 95% CI is around -43% in col 2.

On this thinness

• It seems like you could show us the data.
• Who are the foreign exporters? How do they differ?

​

• I’m not against thin data environments
• But, it is a setting where more transparency about what drives the result would be useful.

​

• Could you drop 1, 2, or 3 foreign exporters from the sample and re-estimate the main results? Would be easy to plot a distribution a la Simonsohn et al (2020)’s specification curve analysis.
• I’ve gotten good reception to this kind of analysis (Cookson 2018; Bellon et al 2021)

Summary

• Interesting paper with important findings.

​

• Most of my concerns are expositional.
• Drop diff-in-diff language.
• Explain the institutional constraints facing foreign/domestic exporters.
• Explore the thinness of the data.

​

• Looking forward to seeing where this project goes.

​