Constrained Poisson pseudo maximum likelihood estimation of structural gravity models

Pfaffermayr, Michael

doi:10.1016/j.inteco.2019.11.014

Cited by 16 publications

(34 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As a methodological contribution, the paper introduces a constrained panel pseudo-Poisson maximum likelihood (PPML) estimation approach, adapting the constrained PPML estimator for cross-sections proposed in Pfaffermayr (2020) to a panel setting. The constrained panel PPML estimator concentrates out countrypair fixed effects and exploits the restrictions imposed by the system of multilateral resistances not only for comparative static analysis, but also for estimation.…”

Section: Introductionmentioning

confidence: 99%

“…Second, under constrained PPML estimation the estimated standard errors of the structural trade cost parameters typically exhibit a much lower bias (downward and of the order of the number of countries) as demonstrated in Pfaffermayr (2020). Moreover, it is straight forward to apply the delta method to derive confidence intervals for counterfactual predictions which might serve as an alternative to computationally intensive bootstrapping procedures.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Trade creation and trade diversion of economic integration agreements revisited: a constrained panel pseudo-maximum likelihood approach

Pfaffermayr

2020

Rev World Econ

Self Cite

View full text Add to dashboard Cite

For PPML estimation of high-dimensional structural gravity panel models it proves useful to exploit the equilibrium restrictions imposed by the system of multilateral resistances. The main advantage of this approach lies in the functional dependence of the parameters of all dummy variables on the structural trade cost parameters. Moreover, the delta method is used to establish confidence intervals of counterfactual changes. Using the constrained panel PPML estimator for a panel of 65 countries in the period 1994–2012 indicates significant trade creation of economic integration agreements with average ranging in between 12.2 and 30.3% in 2012. Results also point to substantial domestic and international trade diversion, where the former dominates the latter.

show abstract

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Trade creation and trade diversion of economic integration agreements revisited: a constrained panel pseudo-maximum likelihood approach

Pfaffermayr

2020

Rev World Econ

Self Cite

View full text Add to dashboard Cite

show abstract

“…However, as only recently the consistency of the model parameter estimates in non‐linear panel models with two types of fixed effects has been shown by Fernández‐Val and Weidner (), we leave this potentially important analysis for future research. See Pfaffermayr () for a recent effort in this direction.…”

mentioning

confidence: 99%

GEPPML: General equilibrium analysis with PPML

2018

View full text Add to dashboard Cite

GEPPML uses a theoretical property that only holds for PPML; however, GEPPML can be implemented with any given trade costs vector. GEPPML has some advantages over alternative methods of measuring gains from trade. Quantitative general equilibrium trade policy analysis imposes both estimation (of key parameters) and computation (of counterfactual equilibria) burdens on the analyst and his readers. A typical counterfactual comparative static exercise using gravity is to change some bilateral friction, for example, remove a tariff and then calculate the effects on trade flows and other variables of interest. The partial effect is based on the estimated bilateral friction, for example, the percentage reduction in buyers' price times the trade elasticity. The general equilibrium impact requires the new counterfactual multilateral resistances, typically solved from the equation system with a nonlinear solver. Our alternative GEPPML method is a more readily accessible way to generate the general equilibrium comparative statics of gravity models. Another benefit is the combination of statistical with economic theoretic intuition in interpreting results. The estimated fixed effects (and their changes) provide traditionally strong fit to the data (under the PPML structure) along with satisfying equilibrium market clearance and budget constraints.We derive our procedure of the comparative statics of gravity models in conditional general equilibrium, the modular trade impact of Head and Mayer (2014), as well as the full general equilibrium impact when endowments are fixed but sellers' prices change (Head & Mayer's, 2014 general equilibrium trade impact), such as in the Ricardian Eaton-Kortum model (Eaton & Kortum, 2002) and the Armington-CES model of Anderson and van Wincoop (2003). Multisector applications beyond the scope of this paper can similarly use GEPPML to calculate Modular Trade Impacts for each sector that nest in any compatibly separable intersectoral general equilibrium production model. This includes most applied general equilibrium models. 1 The second contribution is a hybrid combination of GEPPML and the "exact hat" algebra calibration methods of Dekle, Kortum (2007, 2008). 2 Our main analysis differs quantitatively from the usual applications of Dekle et al. (2007, 2008) in basing calculations on fitted (predicted) trade flows rather than observed trade flows, with the presumed advantage of controlling for measurement error in the trade flow data. We propose a hybrid "estibrating" procedure 3 that obtains estimates of some key trade cost components but also treats the gravity error term as a component of trade costs, thus calibrating the trade cost vector to fit the data. We demonstrate that our methods deliver results identical to those from Dekle et al. (2007Dekle et al. ( , 2008 when fitted trade values are replaced with observed values. An important difference between the two procedures is that GEPPML delivers trade cost response elasticities for specific variables/policies of interest and with the ...

show abstract

“…The MC results are focussed on the bias and consistency of the coefficient estimates, but did not examine the efficiency of the estimators or if and how strongly the estimated standard errors are affected by the incidental parameter problem. Future research could build upon the work of Pfaffermayr (), who showed that his derived constrained PPML—which leads to identical point estimates as a dummy variable PPML in cases of no missing data—yields asymptotically unbiased standard errors. Furthermore, in the DGP employed in this paper, we made ad hoc assumptions on the distribution of the time‐invariant directional pairwise trade costs, since there is no clear empirical benchmark for the size of those effects.…”

Section: Discussionmentioning

confidence: 99%

“…Regarding panel data applications, Fernández‐Vala and Weidner () introduced a correction for non‐linear models, including a Poisson model, with two fixed effects. Furthermore, Pfaffermayr () showed that his proposed constrained PPML, that exploits the theoretical SGM constraints instead of including exporter–period and importer–period dummy variables as fixed effects, is not affected by the incidental parameter problem.…”

Section: Nondiscriminatory Trade Policies In Gravity Modelsmentioning

confidence: 99%

Non‐discriminatory Trade Policies in Panel Structural Gravity Models: Evidence from Monte Carlo Simulations

Sellner

2019

Rev International Economics

View full text Add to dashboard Cite

This paper provides Monte Carlo (MC) simulation evidence on the performance of methods used for identifying the effects of nondiscriminatory trade policy (NDTP) variables in panel structural gravity models. The benchmarked methods include a fixed effect (FE) estimator that utilizes data on intra national trade flows, the bonus‐vetus (BV) and the two‐stage fixed effect (FE‐2S) estimator. The results indicate that only the FE estimates are unbiased and consistent under very general assumptions of the data generating process. The favourable asymptotic properties of the FE estimator unfold as the number of period T increases.

show abstract

Constrained Poisson pseudo maximum likelihood estimation of structural gravity models

Cited by 16 publications

References 42 publications

Trade creation and trade diversion of economic integration agreements revisited: a constrained panel pseudo-maximum likelihood approach

Trade creation and trade diversion of economic integration agreements revisited: a constrained panel pseudo-maximum likelihood approach

GEPPML: General equilibrium analysis with PPML

Non‐discriminatory Trade Policies in Panel Structural Gravity Models: Evidence from Monte Carlo Simulations

Contact Info

Product

Resources

About