Acemoglu, Johnson, Robinson (2005)

Return to BPP Field Exam Papers 2012

Empirical Questions:

What is the author's topic and hypothesis?

The author attempts to show that Atlantic trade contributed to European growth through an indirect instutional channel as well as via its more obvious direct effects.

• Atlantic trade generated large profits for comercial interests in favor of institutional changes in countries that met two crucial criteria:

1. easy access to the Atlantic
2. Nonabsolutisit initial insitutions

These profits swung the balence of power away from the monarchy and induced significant reforms in political institutions, which paved the way for more secure property rights.

How does the author test the hypothesis?

Atlantic Trading and differential Growth

The authors test the the idea that WE growth after 1500 was primarily due to growth in countries involved with Atlantic trading or with high potential for Atlantic trading.

Specification 1 (results in table 2):

[math]u_{jt} = d_{t} + \delta_{j} + \sum\limits_{t\gt =1600} \alpha_{t} \cdot WE_{j} \cdot d_{t} + \sum\limits_{t\gt =1500} \beta_{t} \cdot PAT_{j} \cdot d_{t}+X_{jt} \cdot \gamma +\epsilon_{jt}[/math]

• [math]u_{jt}[/math] is urbanization in country j at time t
• WE is dummy for Western Europe
• d is the year effects
• [math]\delta_{j}[/math] is country effects
• X is other covariate
• PAT is potential for Atlantic Trade

Reults of Specification 1:

• when Interation term added, WE covariate becomes small and non-significant. Demonstrating that the interaction is what is important.
• Run cols 8-10 with Atlantic costline instead of Atlantic Trader, as we are concerned with Atlantic trader being an expost outcome and endogenous.
• Iteration only becomes significant after dates 1600 when using AT dummuy, and 1750 when using Atlantic costline instead of expost AT result.

Specification 2 repeats specification 1 using a more restricted model, as we force differential growth to be related to log of aggregate volume of atlantic trade in time t and country j instead of simply being an Atlantic trader. Results are nearly identical to Specification 1.

The Role of Initial Institutions

The authors now test the part of their hypothesis that it was predominantly societies with less absolutist initial institutions that took advantage of opportunities offered by Atlantic trade.

They estimate the following model:

[math]u_{jt} = d_{t} + \delta_{j} + \sum\limits_{t\gt =1600} \alpha_{t} \cdot WE_{j} \cdot d_{t} + \beta \cdot ln AT_{t} \cdot PAT_{j} + \sum\limits_{t\gt =1500} \gamma_{t} \cdot I_{j,1415} \cdot d_{t} + \eta \cdot ln AT_{t} \cdot PAT_{j} \cdot I_{j,1415} + \epsilon_{jt}[/math]

• [math]I_{j,1415}[/math] is the country j's initial institutions calculated as the average of constraint on the executive in 1400 and 1500.

• The [math]\gamma_{t} \cdot I_{j,1415} \cdot d_{t}[/math] term allows for any differential economic trends related to differences in initial institutions that would apply even without access to Atlantic trading.

• [math]ln AT_{t} \cdot PAT_{j}[/math] term measures the effect of Atlantic trade for a given level of institutions.

• [math]ln AT_{t} \cdot PAT_{j} \cdot I_{j,1415}[/math] tests the hypothesis of interest A significant [math]/eta[/math] implies that there were differential paths taken by countries with different initial institutions, but divergent relates significantly to whether they took advantage to Atlantic trade.

• Results of Specification (3):[math]ln AT_{t} \cdot PAT_{j}[/math] is significant by itself and with [math]\gamma_{t} \cdot I_{j,1415} \cdot d_{t}[/math]. However, when [math]ln AT_{t} \cdot PAT_{j} \cdot I_{j,1415}[/math] is entered, it is the only significant term.

How does the author rule out alternative hypotheses?

To check the robustness of their results, the authors run specification (2) with a number of additional covariates.

• Add covariate for Protestant Country interacted with year dummies -- either insignificant or marginally significant.
  • when outcome is ln GDP/capita , Religion dummies are significant, but little impact on differential growth.
• Add average number of years at war variable as many social-scientists via war as an important factor in state building.
  • Non-significant and no effect on differential growth
• Add dummy for Roman heritage. As many view roots from the roman empire as significant determinant of rise of europe
  • Insignificant and no effect on differential growth patterns.
• Add interactions between distance from the equator and dates. This is to rule out alternative explanation that economic activity was simply shifting away from the south of Europe and towards the north Atlantic states anyways.
  • These are typically insignificant