Don’t

Assume that any trade question fits gravity

Shoehorn an interesting question into gravity

Pursue an ill-defined research question

Do

Define your research question in detail

Match research question, data, and method

Focus on bilateral trade costs

Distinguish estimation from simulation

Don’t

Estimate an intuitive model

Play fast and loose with theory’s implications

Ignore what theory says about data

Ignore what theory says about variables

Ignore what theory says about estimation

Ignore what theory says about simulation

Do

Use “structural” or “theoretical” gravity models

Choose one of the broad class of standard structural models

Note symmetries in estimation & simulation of standard models

Exercise care when developing theories

Don’t

Log linearize

Use OLS

Trust working papers that "improve" on PPML

Simplify fixed effects

Do

Use PPML

Value consistency over efficiency

Use fixed effects dictated by theory

Use ppmlhdfe in Stata

Use glmhdfe in R

Consider including domestic shipments

Don’t

Over-interpret parameter estimates

Calculate projections

Ignore 3rd country effects

Do

Carefully define policy shocks

Conduct theory-based simulations

Use ge_gravity in Stata or R

State limits and assumptions

Explain that results are not forecasts

Don’t

Do theory unless you’re a theorist

Over-interpret standard models

Shoehorn extensions into standard theory

Do

Think about multi-sector models

Think about dynamic models

Think about gravity with unemployment

Think about gravity with comparative advantage

Use domestic shipments to identify MFN policy effects

Make full use of replication code