Notes

These are my personal notes on set of papers that I read for reference.

Research general

Mahoney (2022): Principles for Combining Descriptive and Model-Based Analysis in Applied Microeconomics Research.
- From descriptive to model-based analysis: Motivate readers from descriptive (showing the variation you are leveraging); Let the descriptive guide your model.
Berk, Harvey, and Hirshleifer (2017): How to write an effective referee report and improve the scientific review process.

Topics: Trade

Eaton and Kortum (2002, 2012): “Technology, geography, and trade”, “Putting Ricardo to work”

Contribution and empirical limitation of the traditional Ricardian model: It provides a foundational rationale for why trade occurs and characterizes the gains from trade regarding comparative advantage and productivity differences. However, this approach faces computational challenges: 1) it leads to discrete equilibrium types that are tedious to compute, and 2) tracking comparative advantage across multiple countries over multiple goods becomes empirically intractable.
Contribution and limitation of Dornbusch, Fischer, and Samuelson (1977): This moves to a multiple-goods setting by employing a continuum of goods, which eliminates the discrete wedges in the equilibrium problem found in the traditional Ricardian model. It also introduces trade costs to rationalize the fact that countries tend to consume more domestically produced goods. However, this framework remains limited to a two-country setting.
Contribution of Eaton and Kortum (2002): They incorporate productivity as realizations of random variables drawn from a Fr'echet distribution. Since this distribution is smooth, it solves the discrete wedge issues in the equilibrium. Furthermore, the model is highly tractable because we only need to track the distribution parameters and the statistics derived from them.
Role of \(A_i\): Governs absolute advantage (a higher value implies that the labor requirement is likely to be lower for any given good).
Role of \(\theta\): Governs comparative advantage (a higher value implies that the variability of the labor requirement is small).

Antras, Fort, and Tintelnot (2017): “The margins of global sourcing: Theory and evidence from US firms”

Understanding the tension between intensive margin (sourcing potential through lower marginal cost) and extensive margin (heterogeneous country-specific fixed cost).

Hsiao (2026): “Coordination and commitment in international climate action: Evidence from palm oil”

This paper develops a dynamic empirical framework to quantify the impact of trade policy as a substitute for weak domestic environmental regulations. While such import tariffs can do an equivalent job of targeting emitters, the effect is dampened if coordination and commitment fall short.

Method: demand estimation

This is on a separate blog post.

Method: Production function estimation

Olley and Pakes (1996): “The dynamics of productivity in the telecommunications equipment industry”

Proably the first paper that formalizes the proxy variable estimation method.
Idea is using investment as a proxy for unobserved productivity. Under certain assumptions, we can invert this to apply control function to partial out the unobserved productivity shock from the estimation process.

Levinsohn and Petrin (2003): “Estimating production functions using inputs to control for unobservables”

This paper is more about solving a practical issue that comes up when employing OP method in data.
Either due to data limitation or reality, there are many cases where investment is 0 for most firms. This means either the strict monotonicity assumption cannot be kept or we need to throw out these observations which seem arbitrary.
LP propose one solution: use intermediate input (which are usually used consistently by firms) and apply similar assumptions.

Ackerberg, Caves, and Frazer (2015): “Identification properties of recent production function estimators”

ACF argues that (at least in principle) OP and LP’s identification strategy suffers from functional dependence problem.
In OP and LP’s case, as labor and material is both flexible input, there is not additional variation left in labor that can be used to identify the coefficient on labor input in the first-stage of the estimation.
Their proposed solution is to adjust the timing assumption of the labor input (from \(t\) to \(t-b\) period) and moving the identification of all the coefficients to second-stage of the estimation.

Gandhi, Navarro, and Rivers (2020): “On the identification of gross output production functions”

Focus on nonparametric identification of gross output production function estimation in the presence of both flexible and nonflexible inputs under the model structure of the proxy variable approach.
The paper shows that at least for gross output production (where material input is in RHS of the equation), previous ACF still suffers from identification issue.
Unless we have additional sources of exogenous variation (e.g. price), proxy variable structure does not suffice to identify the gross output production function.
One solution posed by GNR is linking the FOC condition for the flexible input to the production function.
However, note that this can be “somewhat” alleviate if we consider value-added function (material is not in the equation) or if we assume Leontieff production function (aka structural valude added). \(Q = \min (\gamma M, F(L, K))\) and cost minimization gives \(Q = \gamma M = F(L,K)\).