Environmental policy performance and its determinants: Application of a three-level random intercept model

Abstract

We propose the use of a three-level random intercept model to measure the degree of environmental policy performance of different countries and to study its determinants. Inspired by the literature on multilevel latent models and Item Response Theory (IRT), this framework treats policy commitment as a latent variable which is estimated conditional on the difficulty of the policy portfolio implemented by each country. We contribute to the study and scoring of environmental and energy policies in three main ways. First, the model results in a ranking of countries which is conditional on the complexity of their chosen policy portfolio. Second, we provide a unified framework in which to construct a policy indicator and to study its determinants through a latent regression approach. The resulting country ranking can thus be cleaned from the effect of economic and institutional observables which affect policy design and implementation. Third, the model estimates parameters which can be used to describe and compare policy portfolios across countries. We apply this methodology to the case of energy efficiency policies in the industrial sectors of 29 EU countries between 2004 and 2011. We conclude by highlighting the future possible applications of this approach, which are not confined to the realm of environmental and energy policy.

Measuring Environmental Policy Stringency: Approaches, Validity, and Impact on Energy Efficiency

“…Building on the IRT framework, Galeotti et al (2014) extend the application of the IRT model by proposing an ordinal logistic three-level random intercept (TLRI) model to assess the performance of countries with respect to environmental policy. 12 The levels considered are "policy instrument i" (level 1), "time t" (level 2) and "country c" (level 3) and allow to use the IRT model in a panel setting.…”

“…Hence, the sum of the third and second level country residuals provides a time-varying country-level indicator which informs on the "ability" of a country c to implement different policy instruments, which is conditional on the intrinsic difficulty of implementing each instrument estimated in the model (captured by k and (β+τ)). Recall that, as explained in Galeotti et al (2014), this measure can be considered a proxy for each country's commitment to more stringent environmental policy.…”

“…Evidence is available upon request. 11 A standard application of IRT is in the realm of testing educational attainment in questionnaires, as difficulty of the various questions in the test and the ability of a student jointly determine the outcome of the test.12 The basic IRT model considers two levels: the item and the subject Galeotti et al (2014). set up a three level random intercept model to take into account a third dimension of their dataset, which is time.…”

“…set up a three level random intercept model to take into account a third dimension of their dataset, which is time. In the TLRI model, the first level is the item (in this specific case, the policy instrument under consideration), the second level is time and the third level is the subject (in this case, the countries) Galeotti et al (2014). use this model to assess the performance of EU countries with respect to energy efficiency policies.…”

“…The score is equal to the number of laws/regulation which were active in each category at each point in time. Hence, in the specific application ofGaleotti et al (2014) the scores provided no information on the stringency of the policies, just on their presence. This differs from the approach used here, which assigns for each country a score which is higher, the higher the stringency of the policy.…”