# Machine Learning for Public Administration Research, with Application to Organizational Reputation

**Authors:** L. Jason Anastasopoulos, Andrew B. Whitford  
**arXiv:** [1805.05409](https://arxiv.org/abs/1805.05409)  
**Date:** 2018-05-11  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/1805.05409v2/

## Key points

- Machine learning methods are increasingly popular in public administration research
- These techniques allow analysis of various types of data such as text and images
- Lack of clarity regarding proper use and potential pitfalls
- Authors present a comprehensive "guide to practice" for machine learning in public administration research
- Machine learning can develop new measures and tap into new sources of data
- Statistical inference and causal inference can be applied using machine learning techniques
- Pitfalls include unvalidated measures and lack of transparency in interpreting outcomes
- Example provided on analyzing tweets from federal agencies to study organizational reputation

## Summary

Machine learning methods have become increasingly popular in the field of public administration, allowing scholars and practitioners to analyze various types of data such as text and images. These techniques provide an opportunity to test existing theories and develop new ones. However, there is a lack of clarity regarding their proper use and potential pitfalls. To address this gap, the authors of this paper present a comprehensive "guide to practice" for machine learning in public administration research. The paper takes a foundational view of machine learning and highlights its potential benefits for enriching public administration research and practice. One key advantage is the ability to develop new measures by leveraging machine learning algorithms. Additionally, these methods enable researchers to tap into new sources of data that were previously inaccessible or underutilized. Another important aspect discussed in the paper is the application of statistical inference and causal inference in a principled manner using machine learning techniques. By employing these methods, researchers can gain valuable insights into complex relationships within public administration systems. However, the authors also acknowledge the pitfalls associated with using machine learning methods. One major concern is the use of unvalidated measures, which can lead to inaccurate results and flawed conclusions. Another challenge lies in interpreting the outcomes generated by machine learning algorithms, as they often lack transparency. To illustrate how machine learning can be applied in practice, the authors provide an example focusing on organizational reputation in federal agencies. They demonstrate how tweets from 13 executive federal agencies can be analyzed using machine learning techniques to gain insights into reputation dynamics. In summary, this paper offers a comprehensive guide for public administration scholars and practitioners on effectively utilizing machine learning methods. It emphasizes their potential benefits for developing new theories, accessing new data sources, conducting statistical inference, and addressing challenges such as unvalidated measures and interpretability issues. The provided example showcases how these techniques can be applied to study organizational reputation in federal agencies using social media data.