Communicating Managed Retreat in California
Communicating Managed Retreat in California
Managed retreat, the planned relocation of homes and structures in response to rising sea levels, has generated a political firestorm in the state of California. Pushed largely by the powerful Coastal Commission, managed retreat's implementation has largely been blocked by local governments due to immense public opposition to the policy. Local governments and some state officials that have attempted to implement the policy -- most famously Pacifica's mayor John Keener and Commissioner Charles Lester -- have been ousted.
The result? Local governments have largely turned their planning focus to sand replenishment or seawalls, two measures largely blocked by the Coastal Commission. Because of these opposing forces, California's coastal policy has reached a stable policy equilibrium: inaction.
Something has to give. I argue that a compromise position on managed retreat needs to arise, and that the biggest barrier to compromise is a fundamental disparity in the ways the public and government think and talk about managed retreat.
My analysis finds that governments and the public do talk about managed retreat in systematically different ways, with the public focusing on themselves and their homes, whereas governments talk about adaptation and the future. However, I find that while the substantive focus of managed retreat conversations differ, governments and the public do not have significant differences in sentiment (the way they feel about managed retreat).
Thus, despite managed retreat's controversial reputation, there is a chance for compromise on the policy, but it will require governments -- and the public -- to rethink the role managed retreat will play in California's future.
Dataset & Collection: I analyze all quotations about managed retreat from members of the public and the government found in news articles, interviews, and government planning documents (like LCPs, SLR vulnerability assessments). I tokenize the data, classify it as either public or government, record the city the speaker is from, and store the results in a personal database (available on request).
Summary statistics:
n= 78
n(Pacifica) = 19
n(Del Mar) = 28
n(Imperial Beach) = 24
1-Grams: This test examines singular words and their usages by the government vs. the public, with Chi^2 test statistics on the x-axis. P-values are not included in this graph, but the 'future,' 'homes,' and 'decades,' observations are the only ones to reach statistical significance at the conventional alpha = 0.05 level. This analysis was done using the family of 'quanteda' packages in R.
2-Grams: This test examines pairs of words words and their usages by the government vs. the public, with Chi^2 test statistics on the x-axis. P-values are not included in this graph, but none of the pairs achieved significance.
Sentiment Analysis: This method uses the 'sentimentr' package in R to conduct a sentiment analysis (see link for more details) on my database. Scores < -1 indicate negative feelings in a text, scores > 1 indicate positive feelings in a text, and scores contained between (-1, 1) indicate neutral sentiment.
The mean sentiment score for the public is -0.01, neither substantively nor significantly different from 0, and the mean sentiment score for the government is 0.001, again indistinguishable from 0. The joint distribution of scores is plotted below, with the public-classified documents in red and government documents in blue.