r-statistics.co by Selva Prabhakaran

About

R vs Stata: The Economist's Complete Guide to Choosing Your Tool

Stata has been the workhorse of empirical economics for decades. R is gaining ground with its free license, visualization power, and expanding econometrics ecosystem. This guide compares both tools through an economist's lens — with a focus on the methods you actually use.

If you're an economics PhD student choosing your primary tool, or a working economist considering whether to add R to your toolkit, this comparison covers everything that matters: econometric capabilities, workflow, cost, and what the job market rewards.

Cost and Access

Factor R Stata
Student license Free $48/year (IC), $98/year (SE)
Perpetual license Free $295 (IC), $595 (SE), $1,195 (MP)
Annual license Free $125 (IC), $255 (SE), $545 (MP)
After graduation Still free Need new license if not perpetual
Coauthors can verify Yes (always) Only if they have Stata

Stata editions: IC (limited variables), SE (standard), MP (multiprocessor). R has no feature restrictions.

Stata advantage: A perpetual SE license at $595 is reasonable for a single researcher. No packages to install for core econometric methods.

R advantage: Free means no barriers. Every coauthor, reviewer, and student can run your code without purchasing anything.

Econometric Methods Comparison

This is what matters most. Here's how the two compare for the methods economists use daily.

Method Stata R Package
OLS regression regress lm()
Robust standard errors vce(robust) sandwich + lmtest
Clustered SEs vce(cluster id) sandwich::vcovCL(), fixest
Fixed effects xtreg, fe / reghdfe fixest::feols(), plm
Instrumental variables ivregress 2sls fixest::feols(), ivreg
Difference-in-differences diff / did_multiplegt did, fixest, did2s
Regression discontinuity rdrobust rdrobust, rddensity
Panel data xtreg, xtabond plm, fixest
Quantile regression qreg quantreg
Survival/duration models stcox, streg survival
Synthetic control synth Synth, tidysynth, augsynth
Propensity score matching psmatch2, teffects MatchIt, cobalt
Event studies eventdd fixest, did2s
Spatial econometrics Limited spdep, spatialreg

The fixest Game-Changer

The fixest package by Laurent Berge has transformed R econometrics. It handles fixed effects, instrumental variables, clustered standard errors, and multi-way fixed effects — all faster than Stata's reghdfe.

library(fixest) # OLS with two-way fixed effects and clustered SEs # Equivalent to Stata: reghdfe y x1 x2, absorb(firm year) cluster(firm) model <- feols(y ~ x1 + x2 | firm + year, data = df, cluster = ~firm) summary(model) # IV regression with fixed effects # Equivalent to Stata: ivregress 2sls y x1 (x2 = z1 z2), absorb(firm) iv_model <- feols(y ~ x1 | firm | x2 ~ z1 + z2, data = df) # Multiple outcomes at once (no Stata equivalent in one line) multi <- feols(c(y1, y2, y3) ~ x1 + x2 | firm + year, data = df) etable(multi) # Beautiful regression table


  






Workflow and Reproducibility



Stata workflow: Write .do files, produce .log files. Well-established in economics. Everyone knows how to read a .do file.



R workflow: Write .R scripts or R Markdown/.qmd documents that combine code, output, and text.







Aspect
Stata
R






Script files
.do files
.R files




Reproducibility
.do + .log files
R Markdown, Quarto (code + output + text)




Version control
Works, but less common
Git/GitHub deeply integrated




Package management
Built-in ssc install
renv for project-specific libraries




Tables to LaTeX
esttab, outreg2
modelsummary, fixest::etable(), stargazer




Figures
Adequate but limited
ggplot2 (publication-quality)







Regression Tables



Both tools can export regression tables to LaTeX, but R's options are more flexible:





  


    
library(modelsummary)

models <- list(
  "OLS" = feols(y ~ x1 + x2, data = df),
  "FE" = feols(y ~ x1 + x2 | firm, data = df),
  "IV" = feols(y ~ x1 | firm | x2 ~ z1, data = df)
)

# One line to produce AER-style table
modelsummary(models, stars = TRUE, output = "latex")


    


      

      

    


    


  






Data Visualization



This is R's biggest advantage over Stata. Economics papers increasingly require high-quality figures, and ggplot2 is unmatched.





  


    
library(ggplot2)
library(fixest)

# Event study plot — common in applied micro
event_model <- feols(y ~ i(time_to_treat, ref = -1) | unit + year, data = df)

# fixest makes this trivial
iplot(event_model,
      xlab = "Time to Treatment",
      ylab = "Coefficient Estimate",
      main = "Event Study: Effect of Policy on Outcome")


    


      

      

    


    


  






Stata's graphics have improved, but customizing them remains tedious compared to ggplot2's layered grammar.



Performance Benchmarks



For large datasets, speed matters. Here's how they compare for common operations:







Operation
Stata (reghdfe)
R (fixest)
Winner






OLS, 1M obs, 2 FE
~5 sec
~1 sec
R




OLS, 10M obs, 2 FE
~60 sec
~8 sec
R




IV, 1M obs, 1 FE
~8 sec
~2 sec
R




Clustered SEs
Fast
Fast
Tie







fixest is often 5-10x faster than reghdfe for high-dimensional fixed effects regressions. This matters when you're running hundreds of specifications.



What Top Economics Departments Use



The trend is clear: R adoption is growing rapidly in economics, while Stata remains the default.



Stata strongholds:



    


  • Most applied microeconomics research
    • 


  • Labor economics, development economics, public finance
    • 


  • The majority of published empirical economics papers
    • 





R adoption areas:



    


  • Macroeconomics and structural estimation
    • 


  • Causal inference research (many new methods are R-first)
    • 


  • Data visualization for publications
    • 


  • Researchers who also do machine learning
    • 


  • Central banks (ECB, Bank of England, Federal Reserve)
    • 





Both: Many economists use both. They do data cleaning and quick regressions in Stata, then produce figures in R.



The Case for Stata



Choose Stata if you:



    


  • Work in an economics department where everyone uses Stata
    • 


  • Need to collaborate with coauthors who only know Stata
    • 


  • Want every core econometric method available without installing packages
    • 


  • Value Stata's official documentation (it's genuinely excellent)
    • 


  • Do primarily applied microeconomics
    • 





The Case for R



Choose R if you:



    


  • Want free software that works everywhere
    • 


  • Need publication-quality graphics
    • 


  • Work with large datasets where performance matters
    • 


  • Want to use modern causal inference methods (many are R-first)
    • 


  • Plan to do machine learning alongside econometrics
    • 


  • Want reproducible research workflows with R Markdown/Quarto
    • 


  • Work at a central bank, think tank, or tech company
    • 





Migration Tips: Stata to R







Stata Command
R Equivalent






use "data.dta"
haven::read_dta("data.dta")




gen x = ...
dplyr::mutate(x = ...)




keep if ...
dplyr::filter(...)




collapse (mean) y, by(group)
`dplyr::group_by(group)
> summarise(y = mean(y))`




merge 1:1
dplyr::left_join()




reshape long/wide
tidyr::pivot_longer() / tidyr::pivot_wider()




regress y x1 x2
lm(y ~ x1 + x2, data = df)




esttab
modelsummary() or fixest::etable()




graph export
ggsave()







FAQ



Q: Can R read Stata .dta files? A: Yes. haven::read_dta() reads all Stata file versions, preserving variable labels and value labels. The labelled package helps work with labeled data in R.



Q: Is fixest really faster than reghdfe? A: Yes, consistently. Benchmarks show fixest is 5-10x faster for high-dimensional fixed effects. It also handles instrumental variables, Poisson regression, and negative binomial models with the same fast engine.



Q: My advisor uses Stata. Should I still learn R? A: Learn both. Use Stata for collaboration with your advisor, and learn R for visualization, reproducibility, and career flexibility. The investment in R will pay off whether you stay in academia or move to industry.



What's Next




        


        
      


      

        

        
© 2016- Selva Prabhakaran.
          This work is licensed under the Creative Commons License.