Grantee Research Project Results
Final Report: Assessing Preferences for Environmental Decisions with Long-Term Consequences
EPA Grant Number: R826611Title: Assessing Preferences for Environmental Decisions with Long-Term Consequences
Investigators: Keller, Robin , Guyse, Jeffery L. , Biswas, Dipayan , Strazzera, Elisabetta , Eppel, Thomas
Institution: University of California - Irvine
EPA Project Officer: Chung, Serena
Project Period: July 1, 1998 through June 30, 2000 (Extended to July 31, 2001)
Project Amount: $181,851
RFA: Decision-Making and Valuation for Environmental Policy (1998) RFA Text | Recipients Lists
Research Category: Environmental Justice
Objective:
The main objectives of the study were to: (1) expand previous research on discount rates for monetary consequences to non-monetary consequences (e.g., environmental or mortality impacts); (2) conduct experimental studies to discover what factors have an impact on monetary and non-monetary preferences for temporal sequences; and (3) cross-validate and generalize the experimental results with information from professional analysts to guide development of procedures that will assist policy makers in the determination of models for the temporal perceptions of different stakeholders in environmental decisions.Summary/Accomplishments (Outputs/Outcomes):
An experimental study, reported in paper #1 on "Valuing Environmental Outcomes: Preferences for Constant or Improving Sequences," was designed and conducted to analyze individuals' preferences for sequences of outcomes over time related to air quality and near-shore ocean water quality. We were particularly interested in whether temporal preferences for monetary outcome streams differ from preferences for sequences of non-monetary outcomes. We compared our results for the environmental sequences with results on preferences for sequences of health and monetary outcomes, (from the California graduate business student participants in this study and previous research conducted by Gretchen Chapman, George Loewenstein, and others). Generally, participants gave significantly lower ratings to environmental and health sequences (with equal means) that worsened over time, relative to the ratings they gave to sequences that either remained the same or improved over time. This pattern is reversed when facing sequences of monetary payments. This preference structure held for both short (5-year) and long (50-year) time horizons, and was confirmed with choice data. In other words, the generally observed preference pattern implies negative discount rates for non-monetary, but positive discount rates for monetary outcome sequences. A relationship between expectations and choices also was found. A model proposed by Loewenstein and Drazen Prelec for the valuation of sequences of outcomes was applied to the current data set and compared to the traditional discounting model. In all cases, the model that incorporated "Gestalt" features of the sequence (slope and uniformity) performed better than the net present value discounting model at predicting the mean ratings for the sequences in the different domains.Thus, we found that people have very different preferences for temporal sequences of environmental outcomes and income outcomes. This suggests that decisions involving monetary and non-monetary outcomes should not be modeled by first, pricing out all outcomes in money and then, discounting the monetary stream backwards.
Another study (paper #2) was conducted to investigate if the preference patterns observed in the air and water quality domains also exist in the "lives saved" and "lives lost" domains. Survey results on individuals' preferences for 3-year temporal sequences of mortality numbers are presented and compared to results from previous research on preferences for sequences of monetary, health, and environmental outcomes. Three anomalies (gain/loss asymmetry, short/long asymmetry, and the absolute magnitude effect) found in previous discounting research using pairwise matching questions, suggested that there might be preference differences when outcomes were framed as gains (lives saved) or losses (lives lost), when sequences begin now (short term) or in 15 years (long term), or when the number of total lives involved over a three-year sequence was small (60) or large (36,000).
California college student participants accessed the survey via a Web site and responded to eight scenarios by first providing a preference rank order for the five options in each scenario, and then providing demographic data. The first four scenarios involved 3-year sequences of numbers of lives lost each year. The scenarios varied by whether the sequences started now or in 15 years, and whether the total number of lives lost over the 3 years was 60 or 36,000. The five options in a scenario were different temporal patterns of the sequences (e.g., -20, -20, -20 is a constant number of 20 lives lost each year). Other patterns were steeply improving, moderately improving, moderately deteriorating, and steeply deteriorating over time. The final four scenarios involved numbers of lives saved over time and had the same structure as the first set of scenarios.
The standard discounting model would require a person preferring to have all the lives saved occur in the first year of a sequence, and no lives saved in the ensuing 2 years. (This assumes utility is linear in the number of lives lost). Similarly, in the case of lost lives, all the lives should be lost in the third year of the sequence. Such sequences are "steeply deteriorating over time." Only 25-33 percent of the subjects ranked such sequences first, when lost lives were involved. When lives were being saved, a larger percentage, from 39-43 percent, but still less than 50 percent, ranked first in the sequence, with all lives saved in the first year. So, fewer than half of the subjects conformed with the standard discounting model. Sizable numbers of subjects first ranked the constant sequence (32-43 percent in the lives lost scenarios and 25-29 percent in the lives saved scenarios). Improving sequences were first ranked by 19-24 percent in the lives lost cases and 23-29 percent in the lives saved cases.
We compare our results with prior results on income, health, air, and ocean water quality. Guyse, Keller, and Eppel (paper #1) found people preferred constant or moderately increasing sequences for their own health and for air and near-shore ocean water quality, but preferred steeply decreasing sequences for monetary income. In a study by Guyse ("Empirical Validity of Discounted Utility for Decisions Involving Sequences of Monetary Outcomes," working paper, California State Polytechnic University, Pomona, CA, 2002), which had the same survey design as in this paper, sequences of monetary outcomes were examined. Compared with our mortality sequences data, significantly more people preferred a steeply or moderately decreasing monetary sequence. Also, significantly fewer people preferred the constant sequence of money outcomes than the constant sequence of mortality outcomes. We also found a Gain/Loss effect similar to what Guyse found for monetary sequences. Participants preferred to spread the outcomes if there were lives lost, and consolidate them if there were lives saved. This preference for consolidation was stronger with a smaller magnitude of money or lives.
Based on these results, we conclude that temporal discounting models may be inappropriate for representing the value people place on mortality, health, and environmental outcomes. Preferences for constant or improving sequences should be considered in policy making, rather than presuming all people have a preference for deteriorating sequences.
We also gathered information from professionals working on environmental decisions with long-term consequences (paper #3). Four prominent practitioners were interviewed to discuss their practical experience with discounting and evaluating long-term consequences. The interviews were structured around a set of questions, ranging from general comments about the use of discounting, to specific questions about what discount factors to use and whether such factors should be dependent on the specific nature of the consequences. While the results of these interviews cannot give us a widely representative view on the practice of discounting, they nevertheless reveal important insights. None of the respondents indicated that discounting ought to drive the analysis of long-term consequences or could point to an application where discounting made a major difference. Everybody agreed that the complexity of consequences and the implications of various modeling assumptions (including discounting) should clearly be explained to all involved. Finally, the respondents warned against oversimplifying the richness of stakeholders' values by aggregating the streams of consequences into one number. Instead, the discussion and evaluation of long-term consequences should focus on a sub-aggregated level and should include extensive sensitivity analyses.
Paper #4, "Examining Predictive Accuracy Among Discounting Models," follows up with our experimental work above, and earlier experimental work by others, that identifies problems in consistency of preferences with any single discount rate. We ask the question: what discount rate should be used if you want to go ahead and use a discounting model? We present a methodology to analyze data from experimental surveys on intertemporal preferences. Focusing on the traditional exponential discounting model and the hyperbolic discounting model, we use different specifications to model experimental data published by Richard Thaler. Standard measures of goodness of prediction are then applied to fitted data to select among alternative specifications. We first presented our approach by applying it to simulated data. Then we presented a procedure for statistical estimation of the sample discount rate, testing four specifications. Such an approach could be used when temporal preference data has been collected from stakeholders and a discounting model is fitted to the data.
Journal Articles on this Report : 2 Displayed | Download in RIS Format
Other project views: | All 22 publications | 2 publications in selected types | All 2 journal articles |
---|
Type | Citation | ||
---|---|---|---|
|
Guyse JL, Keller LR, Eppel T. Valuing environmental outcomes: preferences for constant or improving sequences. Organizational Behavior and Human Decision Processes 2002;87(2):253-277. |
R826611 (Final) |
Exit Exit |
|
Keller LR, Strazzera E. Examining predictive accuracy among discounting models. Journal of Risk and Uncertainty 2002;24(2):143-160. |
R826611 (2000) R826611 (Final) |
Exit |
Supplemental Keywords:
discounted utility theory, decision analysis, environmental policy, contingent valuation, nonmarket valuation, decisionmaking, psychological, preferences, survey, modeling, public policy, ambient air, marine water, measurement methods, social science., RFA, Scientific Discipline, Economic, Social, & Behavioral Science Research Program, Ecology and Ecosystems, decision-making, Ecological Risk Assessment, Social Science, Economics & Decision Making, ecosystem valuation, multi-objective decision making, policy analysis, social psychology, community involvement, social impact analysis, decision analysis, decision making, incentives, environmental values, multi-attribute utility, non-market valuation, preference formation, environmental policy, long-term consequences, public values, discount rates, public policy, stated preference, benefits assessment, multi-criteria decision analysisRelevant Websites:
http://www.gsm.uci.edu/~keller
Exit
http://yosemite.epa.gov/ee/epa/eerm.nsf/vwRepNumLookup/EE-0464?OpenDocument
Exit
Progress and Final Reports:
Original AbstractThe perspectives, information and conclusions conveyed in research project abstracts, progress reports, final reports, journal abstracts and journal publications convey the viewpoints of the principal investigator and may not represent the views and policies of ORD and EPA. Conclusions drawn by the principal investigators have not been reviewed by the Agency.