In this paper I examine how the form and structure of collaborative management relates to environmental outcomes. In what follows, I briefly describe the theoretical rationale for this research. I then describe my sample frame and the data used to conduct this analysis. Finally, I detail the multilevel statistical model used to test how specific collaborative management policy characteristics at both the watershed and state level relate to water quality and watershed health.

Rationale

One of the most frequent applications of government-supported collaboration is in watershed management, largely because the complexity of watershed management and the diversity of relevant stakeholders. However, there is little empirical evidence regarding whether collaborative management actually improves the environmental condition of watersheds. In order to efficiently use public resources and maximize environmental benefits, it is important to base collaborative management choices on demonstrated impacts. Thus, in this paper I examine how these specific policy design and implementation decisions affect the environmental impact of collaborative watershed management.

Research Design

I use watershed condition data from a representative sample of 600 large rivers sampled in the eastern United States as part of the EPA’s 2008-2009 National Rivers and Streams Assessment (NRSA). I then collect data and code specific variables that represent basic group characteristics (e.g., group size, year of formation) as well as variables used to test the following hypotheses: 

H1: An increased level of collaborative watershed management group responsibility (whether group is designated an information sharing forum, a planning group, or a policy implementation body) results in better water quality outcomes.

H2: Increased plan and objective formalization for collaborative watershed management results in better water quality outcomes

H3: Inclusiveness and diverse representation in collaborative watershed management results in better water quality outcomes.

I employ a multilevel model that accounts for variation at the watershed and state level. At the first level of the model I estimate a water quality index score  for individual river i in state j:

Υ_ji = α_j + β_jC_ji + Σ_kη_kCollab_kjiC_ji + Σ_lδ_lRivChar_lji + Σ_lγ_lRivChar_ljiC_ji + Σ_pτ_pStress_pji + Σ_pρ_pStress_pjiC_ji + ε_ji

where α_j represents the conditional state mean estimate for non-collaborative watersheds; β_j estimates the conditional impact of collaboration by state (C_ji is a binary indicator of collaborative watershed management for river i in state j); RivChar_lji represents a vector of river characteristics (1 to l); Collab_kji represents a vector of collaborative management characteristics (1 to k); Stress_pji represents a vector of watershed stressors (1 to p); and ε_ji represents the river random error term.

 The conditional state impact is the dependent variable for the second model level:

 β_j = β₀ + Σ_mπ_mStateChar_mj + μ_j

in which equals the effect of state characteristic (1 to m); and μ_j equals the state random error term. A third regression term (also at the state level) models the conditional non-collaborative outcome at the state level as a function of the same vector of state characteristics:

α_j = α₀ + Σ_mλ_mStateChar_mj + ν_j

This multilevel model will test the hypotheses above by examining the coefficient estimates (η_k) for each coded collaborative watershed management characteristic (1 to k). Additionally, I test interactions between management choices and specific watershed stressors to examine context-specific implications.