Correlation Does Not Imply Causation: Essays in Causal Inference

Research output: ThesisPh.D. thesis

49 Downloads (Pure)


The present thesis focuses on improving policy recommendations by providing both newfound theoretical and empirical insights within the field of causal inference. The first two chapters address gaps in the instrumental variable methodology, while the third chapter provides an interesting application of the difference-in-difference (DID) approach to evaluate the effectiveness of the Danish fat tax. Chapter 1 proposes the limited monotonicity (LiM) assumption, which is generally weaker than other forms of monotonicity. Under this assumption, it provides an identification result for the combined complier local average treatment effect (CC-LATE), a novel parameter for a large complier population that has intuitive weights. Chapter 2 considers identification of causal effects of nonbinary, discrete treatments using multiple instruments and including covariates. It presents novel insights into the two-stage least squares methodology and proposes the combined complier average causal response (CC-ACR), a parameter with an intuitive interpretation, under the LiM assumption. It also develops a novel test for detecting local violations of LiM. Chapter 3 employs a doubly-robust DID approach to examine the causal effects of the Danish fat tax on consumer behavior. It finds significant effect heterogeneity across products, with consumption of some products declining during the taxed period and expenditure on others increasing. The brief implementation of the tax led to an overall increase in consumption for certain products in the post-tax period, raising concerns about unintended consequences of the tax. Overall, the thesis makes significant contributions to the fields of causal inference and public health policy. The CC-LATE and CC-ACR parameters are valuable tools for researchers and policymakers alike, and the findings on the Danish fat tax provide important insights for designing effective nutrition policies.
Original languageEnglish
Awarding Institution
  • University of Southern Denmark
  • Mellace, Giovanni, Principal supervisor
  • Dahl, Christian Møller, Co-supervisor
Publication statusPublished - 23. Jan 2024


Dive into the research topics of 'Correlation Does Not Imply Causation: Essays in Causal Inference'. Together they form a unique fingerprint.

Cite this