Hey there. I’m sorry that the second post in this column came so late. It has been half a year since the last time I updated my website. I’ve been busy preparing for my English test, writing papers and some other mundane things. Since I’m a pragmatic reader, most of my reading is not profound and comprehensive enough for me to write something about.
But this paper named “Organized Crime and Economic Growth: Evidence from Municipalities Infiltrated by the Mafia”, published in American Economic Review, is special. I found this paper when I was looking for some reference about how to clarify the mechanism in a DID research. And I found this paper extremely detailed and exquisite. So I want to write some notes about its mechanism section.
Cite this paper：Fenizia, Alessandra, and Raffaele Saggio. 2024. “Organized Crime and Economic Growth: Evidence from Municipalities Infiltrated by the Mafia.” American Economic Review 114 (7): 2171–2200.DOI: 10.1257/aer.20221687
Background:
This paper is a good example for using DID (Difference-in-Differences) method. It is well-known that DID is a causal inference method frequently employed to verify the causal relationship between a certain policy (independent variable X) and a potential outcome (dependent variable Y). This paper, as I concluded, had set:
X: city council dismissals (CCD) in Italy;
Y: the economic growth.
In simple terms, the CCD policy involves the central government dissolving municipal institutions in cases where local governments are suspected of being infiltrated by the Mafia. A committee is then appointed to manage the city with full legislative and executive authority for approximately two years. The central government utilizes the CCD policy to regain control over regions plagued by severe corruption and where the mafia effectively governs.
In the first half of the article, the author uses the CCD policy as an experiment to demonstrate that this policy can promote economic growth. However, the author’s true intention is not merely to prove the effectiveness of this policy but to generalize the findings to a broader conclusion, namely that “organized crime hinders economic development.” This is precisely the issue that needs to be addressed in the mechanism section.
The purpose of this section is to demonstrate that the reason CCD can promote economic growth lies in its crackdown on organized crime. Thus, the core of the authors’ argument is to prove:
① The following mechanisms remain unaffected after the CCD implementation: government revenue, fund allocation, and bureaucratic experience.
② Organized crime is weakened following the CCD implementation.
Falsify irrelevant mechanisms:
Although the policy’s objective is to combat crime, its implementation may also generate various additional effects. If the economic development outcomes are actually driven by these side effects, then the causal relationship between “organized crime and economic development” that the paper aims to prove cannot be substantiated!
Therefore, the author identifies three potential channels that could threaten the paper’s conclusions, all of which are straightforward to understand:
①Government Revenues: CCDs may induce growth if increased government transfers induce a stimulus effect via increased spending;
②Structure of government expenditure: external commissioners or newly elected politicians may direct funds to programs that generate employment effects;
③More experienced bureaucrats: better-managed local government by experienced bureaucrats may itself generate economic growth independent of the effects on the local Mafia.
For ① and ②, the authors use government revenue and fund allocation as the dependent variables and employ the DID model to test whether X significantly affects M. The results are not significant. Since the CCD policy did not affect government revenue, it cannot be the case that the economic growth was driven by an increase in government revenue.
For ③, the authors employed a heterogeneity analysis approach. In some regions, the implementation of the CCD policy was due to reasons such as collective government resignations or unexpected deaths of officials, rather than mafia infiltration. In these cases, the central government also dispatched bureaucrats to temporarily ensure regional control until the next election. Both mafia-related and non-mafia-related CCD implementations were influenced by experienced bureaucrats. By comparing the differences between these two sample groups, it is possible to determine whether economic growth was driven by bureaucratic experience. The results show that the economic effects of non-mafia CCDs were only 20% of those of mafia-related CCDs. Therefore, significant economic effects are only generated when CCDs target mafia-induced infiltration.
Prove the relevant mechanisms:
Weakening the Mafia: How to demonstrate that CCD has indeed weakened the mafia? Select several indicators that reflect this fact:
① Election Results: The mafia promotes patriarchal societal norms and leverages its influence to manipulate elections, making it more likely for men to be elected. Under CCD, without significantly altering the characteristics of candidates, the probability of electing women, younger individuals, those with higher education levels, and first-time candidates (traits associated with lower corruption) has significantly increased.
② Collusion Between Politics and Business: The mafia’s economic success largely depends on its ability to create and maintain monopolies, which often rely on political connections. The proportion of politicians holding both corporate and government positions decreased by 5%, and the likelihood of dismissed mayors or deputy mayors serving on corporate boards dropped by 80%.
③ Industries at High Risk of Mafia Infiltration: If the mafia is weakened, industries prone to mafia monopolies should see an increase in legitimate businesses. CCD led to a nearly 20% rise in the number of businesses in mafia-dominated industries, while having no significant impact on non-mafia industries. Employment numbers also increased by a similar margin.
④ Mafia-Linked Businesses: Local firms that secured public procurement contracts before CCD might have done so due to political corruption with authorities. After CCD, these businesses experienced declines in both employment and value added per capita.
Conclusion:
The CCD did not boost regional income growth or alter the structure of expenditures, and the role of experienced bureaucrats introduced by the policy is insufficient to fully explain the economic growth effects.
However, the CCD did indeed weaken local organized crime, as evidenced by political characteristics aligning more closely with those of low-corruption governments and a decline in the mafia’s control over monopolized industries, among other factors.
Therefore, the weakening of the mafia by the CCD is the mechanism driving the economic effects. This demonstrates that curbing organized crime can promote economic growth, and that organized crime acts as a hindrance to economic development.
Comments:
This article has reshaped my understanding of mechanism testing, primarily in two aspects:
First, mechanism testing should serve to substantiate the core causal relationship argued in the paper. I have always understood that DID papers should not merely focus on assessing the impact of a specific policy but rather on evaluating the broader category of measures that the policy represents. Only then can the research hold practical significance. However, I had rarely considered whether the policy I was using truly exemplified the category of measures I intended to study—even if the policy was originally designed to achieve such objectives.
Second, top-tier papers do not require complex methodologies, nor do they need to rigidly follow a fixed formula. In Chinese DID papers, we often encounter models such as mediation effects and moderation effects in the mechanism testing section. Authors attempt to use these models to demonstrate that their independent variable X influences the dependent variable Y through several pathways. The application of this methodology may appear rigorous, but in reality, it is highly mechanical. Every social science study should be highly contextualized and differentiated, requiring us to delve deeply into discerning truth from falsehood and meticulously unravel the causal relationships.
The most valuable aspect of this article lies in its ability to successfully address a problem that plagues countries worldwide through a case study of a single policy implemented by one nation. Its methodology and research approach are both meticulously crafted, making it worthy of further study.
read on 13/3/2025
