jeff lazo societal impacts program was*is workshop boulder, co july 2006 some economics to make a...
TRANSCRIPT
0101
Jeff LazoJeff Lazo
Societal Impacts ProgramSocietal Impacts Program
WAS*IS WorkshopWAS*IS Workshop
Boulder, COBoulder, CO
July 2006July 2006
Some Economics Some Economics to Make a Wiser to Make a Wiser WAS*ISWAS*ISerer
0101
ec·o·nom·ics (ĕkec·o·nom·ics (ĕk''ə-nŏm'ĭks, ēə-nŏm'ĭks, ē''kə-) kə-) n. n. The social science that deals with the production, The social science that deals with the production,
distribution, and consumption of goods and services distribution, and consumption of goods and services and with the theory and management of economies and with the theory and management of economies or economic systems.or economic systems.
http://www.answers.com/topic/economicshttp://www.answers.com/topic/economics
Lionel RobbinsLionel Robbins 1932: "the science which studies 1932: "the science which studies human behavior as a relation between scarce means human behavior as a relation between scarce means having alternative uses."having alternative uses."
http://en.wikipedia.org/wiki/Economicshttp://en.wikipedia.org/wiki/Economics
study of the allocation of scarce resources in light of study of the allocation of scarce resources in light of unlimited wantsunlimited wants
EconomicsEconomics
0101
ObjectivesObjectives exposed to basic concepts of economicsexposed to basic concepts of economics exposed to basic methods of economicsexposed to basic methods of economics discuss some applicationsdiscuss some applications
Reality checkReality check can’t teach economics in one hourcan’t teach economics in one hour what economics do you need to know?what economics do you need to know? some people think economists think differently some people think economists think differently
than other people – that is NOT truethan other people – that is NOT true other people think differently than other people think differently than
economistseconomists
EconomicsEconomics
Why Value Forecasts?Why Value Forecasts?
1.1. program justificationprogram justification benefit-cost analysisbenefit-cost analysis
2.2. program evaluationprogram evaluation
3.3. guidance for research investmentguidance for research investment any cases of true comparative any cases of true comparative
analysis?analysis?
4.4. inform users of forecast benefitsinform users of forecast benefits
5.5. developing end-to-end-to-end developing end-to-end-to-end forecast and warning systemforecast and warning system
What Should Be Valued?What Should Be Valued?
Weather impactsWeather impacts
Dutton - $3T USDutton - $3T US ForecastsForecasts Improved forecastsImproved forecasts Research to improve forecastsResearch to improve forecasts How forecasts are usedHow forecasts are used
What Should Be Valued?What Should Be Valued?
Forecasts
Value
Integrate forecasting and valuation- Meteorology- Economics
What Should Be Valued?What Should Be Valued?Weather
Observation
Forecast
Communication
Perception
Use
Value
Meteorology Risk communication
Marketing
Psychology Anthropology Sociology
Geography Risk perceptions Economics
Some Basic Some Basic Economics: Economics:
Value Theory and Value Theory and Topics in ValuationTopics in Valuation
Econ 101Econ 101
TOPICSTOPICS value theoryvalue theory consumers and producersconsumers and producers supply and demandsupply and demand markets and pricesmarkets and prices consumer surplusconsumer surplus producer surplusproducer surplus net societal welfarenet societal welfare market failuresmarket failures
Econ 101Econ 101
0101
What is Value?What is Value?
Nelson and Winter QJE
ForecastForecast
FrostFrost No FrostNo Frost
ActioActionn
ProtectProtect -C (lost -C (lost cost)cost)
-C (lost cost)-C (lost cost)
Don’t Don’t ProtectProtect
-L (lost -L (lost crop)crop)
00
Why ask “What is Value?”Why ask “What is Value?”• Ensure that “economic value” is valid
economics• Look at broader approach to economic
valuation
What is Value?What is Value?Market FailuresMarket Failures
Public goods
Market power
Externalities
Information
What is Value? What is Value? Topics: Public Goods Topics: Public Goods
What is the price (i.e., value) of weather forecasts?
Weather forecast characteristics• Non-rival• Non-exclusive
Problems of public goods• No observable price information• No provision by private markets
Weather forecasts as “quasi-public goods”?
What is Value? What is Value? Topics: Time Topics: Time
PeriodBenefit
s Costs
0 0.00 100.00
5 60.00 0.00
10 60.00 0.00
120.00 100.00
Discounting
Discounted Benefits
Discounted Costs
0.00 100.00
47.02 0.00
36.83 0.00
83.85 100.00Net Benefit
20.00Net Benefit -16.15
5% Rate of Time Preference
What is Value?What is Value?Topics: VSLTopics: VSL
Value of Statistical Life (VSL)
• 1,000,000 people each willing to pay $50 a year for a program to reduce the chance of death by 1 in 100,000 per year (say from 20 in 100,000 to 19 in 100,000 each year)
• Means that the group is WTP $50,000,000 to prevent 10 deaths
• VSL = $50,000,000/10 deaths = $5,000,000
What is Value?What is Value?Topics: Benefits TransferTopics: Benefits Transfer
Application of results from one study for a different analysis context.
Same commodity being valued?same baseline? same outcome?
No No ForecastForecast
ClimatologClimatologyy
CurrentCurrent ImprovedImproved PerfecPerfectt
Adjusting for:date of study – changes in prices (inflation)changes in preferenceincome differencesavailability of substitutes and complementsother significant determinants of value
Evaluation of the Evaluation of the Sensitivity of U.S. Sensitivity of U.S. Economic Sectors to Economic Sectors to Weather Weather ((OUSSSAOUSSSA))
Jeffrey K. Lazo – NCARPete Larsen – NCAR / Cornell UniversityMegan Harrod – Stratus ConsultingDonald Waldman – University of Colorado
Purpose: Assess sensitivity of US economic sectors to weather variability
OutlineOutline
MotivationMotivation ConceptConcept What is Economic Sensitivity? What is Economic Sensitivity? Data and ModelingData and Modeling ResultsResults ConclusionsConclusions
Dutton – BAMS – September 2002
“. . . one-third of the private industry activities, representing annual revenues of some $3 trillion, have some degree of weather and climate risk.
This represents a large market for atmospheric information . . . “
Conceptual ApproachConceptual ApproachModel Building:Model Building: using historical using historical economic and weather data, we model economic and weather data, we model the relationship between economic the relationship between economic output in 11 sectors, economic inputs, output in 11 sectors, economic inputs, and weather and weather variabilityand weather and weather variability
Capital
Labor
Energy
Temperature
Precipitation? Gross
StateProduct
Conceptual ApproachConceptual Approach
Sensitivity Analysis:Sensitivity Analysis: Using these Using these models, we then hold the economic models, we then hold the economic inputs constant, and use 70 years of inputs constant, and use 70 years of weather data to see how economic weather data to see how economic output varies as a result of variation in output varies as a result of variation in weatherweather
Capital
Labor
Energy
Temperature
Precipitation
Gross State
Product
Define “Sensitive”Define “Sensitive” No single correct definition
Characteristics of a meaningful approach consistent with economic theory amenable to empirical examination provide meaningful information
about economic impacts of Wx
What is Weather What is Weather Sensitivity?Sensitivity?
P$
Q
S(K0, L0, E0;W
0)
D(W0)
P*
Q*
D(W1)
S(K0, L0, E0; W1)
Q1
P1
Change in GSPGSP
Issues?Issues?
•Weather or climate?
•Sensitivity or something else?
Super SectorsSuper SectorsSector Billions (2000$)
Wholesale Trade 592Retail Trade 662
Transportation 302Utilities 189
Communications 458Agriculture 98
FIRE 1,931Manufacturing 1,426Construction 436
Mining 121Services 675
Total Private Sector 6,890Government 1,135
Total 8,026
Economic ModelingEconomic Modeling
ij WL K Er t
ijt ijt ij it i tijt jtQ A e L K E W
( , , )Q f L K E
ln( ) ln( ) ln( ) ln( ) ln( )
1 1ln( ) ln( ) ln( ) ln( ) ln( )
2 2ln( ) ln( ) ln( ) ln( ) ln( ) ln( )
1ln( ) ln( ) ln( ) ln
2
ijt ijt L ijt K ijt E ijt
it it it LL ijt ijtW WW
LK ijt ijt LE ijt ijt ijt itLW
KK ijt ijt KE ijt
Q A t L K E
W W W L L
L K L E L W
K K K
( )
1ln( ) ln( ) ln( ) ln( ) ln( ) ln( )
2
ijt
ijt it EE ijt ijt ijt it ijtKW EW
E
K W E E E W
Translog FunctionTranslog Function
Weather “Sensitivity”Weather “Sensitivity”
ln ln ln lnWQ A rt LL
W
lnln W
QW
n l ln ,l n , L W WQ L LL
W W
, caused b yQ WQ Q
Economic DataEconomic DataEconomic DataEconomic Data - - state x year x sector
Gross State Product Gross State Product (dependent variable)
Production InputsProduction Inputs Capital (K) - dollarsCapital (K) - dollars Labor (L) - hoursLabor (L) - hours Energy (E) – BTUsEnergy (E) – BTUs
Weather Data Weather Data - - state x year
Temperature VariabilityTemperature Variability CDD : Cooling Degree Days: (T - 65) on a given day HDD : Heating Degree Days: (65 - T) on a given day
PrecipitationPrecipitation P_Tot: Precipitation Total (per square mile)P_Tot: Precipitation Total (per square mile) P_Std: Precipitation Standard DeviationP_Std: Precipitation Standard Deviation
i = state 48j = sector 11t = year 1977-2000 = 24 years48 x 11 x 24 = 12,672 “observations”
Temperature Weather Temperature Weather InputsInputs
CDD: Defined as (T - 65) = daily CDD, where T is daily Average Temperature (F). If T is less than 65 degrees F, CDD=0.
HDD: Defined as (65 - T) = daily HDD, where T is daily Average Temperature (F). If T is greater than 65 degrees F, HDD=0.
Average (Mean) Temperature of the day : (High Temperature + Low Temperature) / 2 ; High and Low Temperature are whole integer values.
http://www.weather2000.com/dd_glossary.html
Econometric MethodsEconometric Methods
• Level data versus per capita
• Panel data – time series – AR(1)
• Heteroskedasticity
• Fixed Effects
• Covariance calculations for marginal effects
Econometric ResultsEconometric Results
Parameter EstimateSig
.
Intercept -12.089 ***
YEAR 0.003 **
Capital 0.672 ***
Labor 0.798 ***
Energy 0.086 **
Heating Degree Days -0.035 ns
Cooling Degree Days -0.068 ***
P_Tot -0.187 ***
P_Std 0.185 ***
Sector: Agriculture
ns = not significant at 10%* 10%, ** 5%, *** 1%
Parameter Estimates from Full Model Parameter Estimates from Full Model RegressionsRegressions
Significance (* = 10%, ** = 5%, *** = 1%, ns = not signficant)DF=1068 for all models
Agric. Wholes Retail FIRE Comm. Utilities Transp. Manf. Constr. Mining Svcs
Inter50.46
ns-1.65ns
-2.08ns
39.98ns
27.08ns
-25.26ns
-28.44ns
-24.78ns
-6.13ns
153.57**
55.72***
YEAR-0.01
***0.02***
-0.01***
0.004**
-0.01***
0.01**
0.01***
0.03***
0.00ns
-0.03***
0.003***
lnKAP
-2.10**
-0.12ns
-0.75ns
7.70***
2.98***
9.21***
4.78***
0.51ns
-9.55***
-5.93***
-2.66***
CDD2.56
*-0.24ns
0.24ns
-3.10**
-0.97ns
2.79ns
-3.52***
1.49ns
-1.43ns
2.50ns
-1.05*
KAP20.14***
0.05***
0.08***
0.02ns
0.02ns
-0.23***
0.07***
0.04ns
0.06***
0.19***
-0.03ns
KAP x HDD
-0.06ns
0.04**
0.09***
-0.07**
0.06**
-0.31***
-0.06ns
0.05ns
0.24***
0.57***
0.15***
Marginal ResponsesMarginal Responses Capital Labor Energy
Sector Marg Eff T-Stat Marg Eff T-Stat Marg Eff T-Stat
Agriculture 1.10 35.02 0.44 8.55 -0.01 -0.14
Communications 1.12 30.08 0.31 12.57 -0.14 -6.23
Construction 0.48 12.40 1.14 52.35 0.12 4.60
FIRE 0.98 32.49 0.39 9.82 -0.20 -6.84
Manufacturing 0.48 5.76 0.62 6.98 0.09 1.71
Mining 1.20 11.86 0.60 9.20 0.10 1.51
Retail Trade 0.91 31.15 0.54 15.94 -0.04 -2.02
Services 0.94 35.85 0.64 18.57 -0.07 -5.53
Transportation 0.94 28.84 0.33 12.21 0.07 1.90
Utilities 1.11 22.57 -0.31 -4.94 -0.03 -0.73
Wholesale Trade 0.50 19.99 0.78 33.01 -0.02 -1.15
Marginal ResponsesMarginal Responses HDD CDD Total Precip Precip
Variance
Sector Marg Eff T-Stat Marg Eff T-Stat Marg Eff T-Stat Marg Eff T-Stat
Agriculture 0.00 -0.05 -0.19 -6.11 0.28 1.89 -0.12 -6.75
Communic. 0.13 3.96 0.06 3.31 0.06 0.36 0.17 16.15
Construct. -0.01 -0.38 0.06 2.85 -0.01 -0.05 0.26 20.84
FIRE 0.15 3.52 0.06 2.70 0.54 3.19 -0.08 -5.60
Manufact. 0.18 1.85 0.02 0.36 0.49 2.34 -0.22 -6.60
Mining 0.25 1.97 0.04 0.57 -3.52 -9.54 1.10 27.44
RetailTrade 0.04 1.75 0.03 2.88 -0.13 -1.32 0.13 18.20
Services 0.04 2.07 0.00 0.29 0.33 4.01 -0.05 -7.72
Transport. -0.03 -0.91 0.01 0.44 -0.15 -0.74 0.15 12.18
Utilities 0.00 0.04 0.08 1.91 -0.59 -1.42 -0.28 -11.59
Wholesale 0.10 4.63 0.02 1.65 -0.19 -1.93 0.02 3.03
Wx Sensitivity AnalysisWx Sensitivity Analysis
Average K, L, E 1996-2000Average K, L, E 1996-2000 Set Year to 2000Set Year to 2000 Historical weather data 1931-2000Historical weather data 1931-2000 Fitted GSP values by sector by state by Fitted GSP values by sector by state by
yearyear 11 sectors11 sectors 48 states48 states 70 “years” fit to 2000 “economic structure”70 “years” fit to 2000 “economic structure”
11 Sector Models:11 Sector Models:Q = Q = f f (K, L, E, W; Year, State)(K, L, E, W; Year, State)
State SensitivityState Sensitivity(Billions $2000)
State SensitivityState Sensitivity(Billions $2000)
State Mean Max Min Range % Range Rank
New York 633.3 679.6 594.0 85.6 13.5% 1
Alabama 92.0 93.9 81.7 12.2 13.3% 2
California 1019.4 1080.5 968.6 111.9 11.0% 3
Wyoming 13.7 14.3 12.8 1.4 10.5% 4
Ohio 312.0 330.6 298.4 32.2 10.3% 5...
......
......
......
Delaware 30.2 30.6 29.6 1.0 3.3% 44
Maine 27.0 27.4 26.5 0.9 3.3% 45
Montana 17.2 17.4 16.9 0.6 3.3% 46
Louisiana 109.5 111.2 107.6 3.6 3.3% 47
Tennessee 141.1 142.8 139.3 3.5 2.5% 48
Sector Sector SensitivitySensitivity (Billions $2000)
Sector SensitivitySector Sensitivity (Billions $2000)
Sector Mean Max Min Range %Range
Agriculture 127.6 134.4 119.0 15.4 12.09%
Wholesale trade 601.5 607.8 594.5 13.3 2.20%
Retail trade 761.5 771.2 753.9 17.3 2.27%
FIRE 1,639.3 1,713.1 1,580.6 132.5 8.08%
Communications 237.3 243.4 232.3 11.1 4.68%
Utilities 212.9 220.8 206.0 14.9 6.98%
Transportation 276.1 280.7 271.0 9.8 3.53%
Manufacturing 1,524.8 1,583.2 1,458.2 125.1 8.20%
Construction 374.5 384.0 366.4 17.7 4.71%
Mining 102.0 108.9 94.2 14.7 14.38%
Services 1,834.9 1,865.4 1,804.9 60.5 3.30%
Sector SensitivitySector Sensitivity (Billions $2000)
Sector Mean Max Min Range %Range
Agriculture 127.6 134.4 119.0 15.4 12.09%Wholesale trade 601.5 607.8 594.5 13.3 2.20%
Retail trade 761.5 771.2 753.9 17.3 2.27%
FIRE 1,639.3 1,713.1 1,580.6 132.5 8.08%
Communications 237.3 243.4 232.3 11.1 4.68%
Utilities 212.9 220.8 206.0 14.9 6.98%
Transportation 276.1 280.7 271.0 9.8 3.53%
Manufacturing 1,524.8 1,583.2 1,458.2 125.1 8.20%
Construction 374.5 384.0 366.4 17.7 4.71%
Mining 102.0 108.9 94.2 14.7 14.38%Services 1,834.9 1,865.4 1,804.9 60.5 3.30%
Sector SensitivitySector Sensitivity (Billions $2000)
Sector Mean Max Min Range %Range
Agriculture 127.6 134.4 119.0 15.4 12.09%
Wholesale trade 601.5 607.8 594.5 13.3 2.20%
Retail trade 761.5 771.2 753.9 17.3 2.27%
FIRE 1,639.3 1,713.1 1,580.6 132.5 8.08%
Communications 237.3 243.4 232.3 11.1 4.68%
Utilities 212.9 220.8 206.0 14.9 6.98%
Transportation 276.1 280.7 271.0 9.8 3.53%
Manufacturing 1,524.8 1,583.2 1,458.2 125.1 8.20%
Construction 374.5 384.0 366.4 17.7 4.71%
Mining 102.0 108.9 94.2 14.7 14.38%
Services 1,834.9 1,865.4 1,804.9 60.5 3.30%
National SensitivityNational Sensitivity (Billions $2000)
Total National 7,692.4 7,554.6 7,813.4 258.7 3.36%
Sector Mean Max Min Range %Range
Agriculture 127.6 134.4 119.0 15.4 12.09%
Wholesale trade 601.5 607.8 594.5 13.3 2.20%
Retail trade 761.5 771.2 753.9 17.3 2.27%
FIRE 1,639.3 1,713.1 1,580.6 132.5 8.08%
Communications 237.3 243.4 232.3 11.1 4.68%
Utilities 212.9 220.8 206.0 14.9 6.98%
Transportation 276.1 280.7 271.0 9.8 3.53%
Manufacturing 1,524.8 1,583.2 1,458.2 125.1 8.20%
Construction 374.5 384.0 366.4 17.7 4.71%
Mining 102.0 108.9 94.2 14.7 14.38%
Services 1,834.9 1,865.4 1,804.9 60.5 3.30%
Future Research (1)Future Research (1)
extend data past 2000 better capital and energy data include “storms” data include forecast skill measure
value of weather forecasts?value of weather forecasts? split supply and demandsplit supply and demand model uncertaintymodel uncertainty
Future Research (2)Future Research (2)
finer spatial scales county level data for a state
finer temporal scales quarterly / monthly economic data
finer sectoral scales 2, 3, or 4 digit sector study
other regions / countries
ConclusionsConclusions
Economically valid analysis Significant impact of weather
• significant regression coefficientssignificant regression coefficients• significant marginal effectssignificant marginal effects
Interpretation of weather sensitivity• upper-bound weather risk measure?upper-bound weather risk measure?• upper-bound measure of value of weather upper-bound measure of value of weather
information?information? 3.4% of annual US economic variability3.4% of annual US economic variability $260B US economic variability related $260B US economic variability related
to weather variabilityto weather variability
0101
Benefits of Investing Benefits of Investing in Weather in Weather Forecasting ResearchForecasting ResearchJeff Lazo, Jennie Rice, Marca Hagenstad
SuperCompSuperComp
Purpose: Assess benefits of buying a new supercomputer for weather forecast research
TOPICSTOPICS Value of investments in Value of investments in
researchresearch Assess value chainAssess value chain Benefit-cost analysisBenefit-cost analysis Benefits transferBenefits transfer Value of statistical lifeValue of statistical life DiscountingDiscounting Sensitivity analysisSensitivity analysis
SuperCompSuperComp
0101
SuperComp Study MethodsSuperComp Study Methods
1.1. Determine potential impact of Determine potential impact of supercomputer on forecast qualitysupercomputer on forecast quality
2.2. Identify potential sectors/users and of Identify potential sectors/users and of improved forecastimproved forecast
3.3. Identify existing benefit studies for Identify existing benefit studies for sectors/userssectors/users
4.4. Quantify probabilities and timing of Quantify probabilities and timing of impactsimpacts
5.5. Develop benefits model for Develop benefits model for aggregating over timeaggregating over time
6.6. Conduct sensitivity analysisConduct sensitivity analysis
0101
Example: “SuperComp”Example: “SuperComp”
New Supercomputer
Improved Environmental
Modeling
Air Force Benefits
DOE Benefits (wind)
Marine Resource Mgt. Benefits
Private Sector Benefits (e.g., highways)
International Benefits
Improved Operational Forecasts (NWS Benefits)
Army Benefits
Aviation Benefits
Retail Benefits
Energy Benefits (temps, wind)
Marine Transportation Benefits
Agriculture Benefits
Total
BenefitsHousehold Benefits
Example: “SuperComp” Example: “SuperComp” Household Benefits of Short Term Weather Household Benefits of Short Term Weather
ForecastsForecasts
Short TermWeatherModeling
Severe WeatherForecasts
TemperatureForecasts
PrecipitationForecasts
WindForecasts
Home ImprovementDecisions
Preparation forSevere Weather
ShoppingDecisions
Commuting /Travel Decisions
Total
Benefits
RecreationDecisions
Cloud CoverForecasts
Stratus Consulting (2002) – stated preference study
Example: “SuperComp”Example: “SuperComp”Agriculture: Annual value of improvement to
perfect information (PI)Apples,
peaches, and pears Alfalfa
Winter wheat
Total - these crops
Value of improvement to
PI per acre of farmland $1,403 $75 $35 $65.19
Acres of farmland 828,460 23,541,000
44,349,000
68,718,460
Value of PI - 100% of land
$1.16 B $1.77 B $1.55 B $4.48 B
Value of PI - 5% of land
$58 M $89 M $77 M $224 M
Example: “SuperComp”Example: “SuperComp”Weather-related fatalities and VSL
estimates (we assume 10% of weather-related fatalities preventable
with perfect information)
Year Fatalities
Fatalities — value
(millions)
1996 540 $3,240
1997 600 $3,600
1998 687 $4,122
1999 908 $5,448
2000 476 $2,856
2001 464 $2,784
2002 540 $3,240
Average annual 602 $3,613a. Calculated as $6 million per fatality.Source: NWS (1996-2002).
Example: “SuperComp”Example: “SuperComp”
Observation
Understanding
Computing
Improvements in Weather Forecasts
NCEP Supercomputing
NHRA
GFDL Supercomputing
HPCS – double current computing capabilities - enable doubling of the spatial and temporal resolutions of environmental models currently run by NOAA, including finite-difference models of the atmosphere and ocean
Example: “SuperComp”Example: “SuperComp”
1.1. Contribution from NHRA – 20%Contribution from NHRA – 20%
2.2. Contribution from computing – 33%Contribution from computing – 33%
3.3. Contribution to forecast improvements Contribution to forecast improvements over 5 year life of NHRA – 75%over 5 year life of NHRA – 75%
20% x 33% x 75% = 5%20% x 33% x 75% = 5%
Example: “SuperComp”Example: “SuperComp”
Financial assumptions for base case present value calculations
Real social discount rate 3%
Decision to purchase supercomputer 2004
First year of operation 2005
Number of years until benefits begin 2
Number of years in which benefits accrue
5
Time horizon for accrued benefits Infinite
Example: “SuperComp”Example: “SuperComp”
Summary of present value of benefits in 2003 (millions, 2002$)
Household sector 69
Orchards, winter wheat, alfalfa 26
Avoided weather-related fatalities 21
0101
What Are Weather What Are Weather Forecasts Worth? Forecasts Worth? Stated Preference Stated Preference Approaches to Approaches to Valuing InformationValuing InformationJeff Lazo, Rebecca Morss, Barb Brown, Stratus Consulting
StormStormPurpose: Assess values to households of ordinary weather forecasts
TOPICSTOPICS Stated preference valuationStated preference valuation Survey developmentSurvey development AnalysisAnalysis
Uses – between city comparisonUses – between city comparison Sources - regressionSources - regression Perceptions – between individual comparisonPerceptions – between individual comparison
Econometrics - regression analysisEconometrics - regression analysis Probit modelProbit model Bivariate probit modelBivariate probit model Combining CV and SC dataCombining CV and SC data
STORMSTORM
0101
Study ObjectiveStudy Objective
• Evaluate benefits to households of Evaluate benefits to households of improvements in weather forecasting improvements in weather forecasting servicesservices
• 104,705,000 households104,705,000 households
• Day-to-day weatherDay-to-day weather
• National Oceanic & Atmospheric National Oceanic & Atmospheric AdministrationAdministration
STATED PREFERENCE
METHODS
REVEALED PREFERENCE
METHODS
NON-MARKET VALUATION METHODS
RATING RANKING CHOICE
Open Ended Contingent Valuation
Referendum Contingent Valuation
Other Choice Based
Methods
Attribute Based Stated ChoiceAdapted from Figure 1: Adamowicz,
Louviere, and Swait. 1998
Survey DevelopmentSurvey Development Atmospheric Science Advisors (ASA)Atmospheric Science Advisors (ASA)
attributes of weather forecastsattributes of weather forecasts current and potential level of attributescurrent and potential level of attributes
Focus groups Focus groups (15 subjects)(15 subjects) One-on-one interviews One-on-one interviews (11 subjects)(11 subjects) Denver Pretest Denver Pretest (84 Subjects)(84 Subjects)
Survey Expert Review PanelSurvey Expert Review Panel North Carolina Focus Groups North Carolina Focus Groups (23 subjects)(23 subjects) Multi-site implementation Multi-site implementation (381 Subjects)(381 Subjects) National random sample National random sample (~1,400 Subjects)(~1,400 Subjects)
Survey LayoutSurvey Layout
IntroductionIntroduction Sources, perceptions and usesSources, perceptions and uses Forecast attributesForecast attributes Value for improved weather forecastsValue for improved weather forecasts
Stated choice - attributes of forecastsStated choice - attributes of forecasts Contingent valuation – demand characteristicsContingent valuation – demand characteristics
Household characteristicsHousehold characteristics Value for Current ForecastsValue for Current Forecasts Severe WeatherSevere Weather
Survey ImplementationSurvey Implementation 9 cities – in-person self-administered9 cities – in-person self-administered written survey - ~25-30 minuteswritten survey - ~25-30 minutes 381 Respondents381 Respondents
Socio-demographicsSocio-demographics
Characteristic Mean
Kruskal-Wallis Test 2
Income (2001$) $49,934 18.84 ** Age 43.7 yrs 13.78 * Education 14.9 yrs 15.12 * Gender 43% males 10.27 How long lived in the area 19.8 yrs 18.29 ** Household size 2.7 6.25 ***, **, * Significant at the 1%, 5%, and 10% respectively
ResultsResults
SourcesSources PerceptionsPerceptions UsesUses Attributes and LevelsAttributes and Levels ValuationValuation
PerceptionsPerceptionsImportance of Weather Forecast Importance of Weather Forecast
CharacteristicCharacteristic
PerceptionsPerceptionsImportance of Weather Forecast Importance of Weather Forecast
CharacteristicCharacteristic
Characteristic Mean SD Kruskal-Wallis
Test, 2 (prob Ho)
Chance of rain, snow, or hail 4.30 0.82 12.44 (0.13)
Amount of rain, snow, or hail 4.02 0.96 21.73 (0.01)
High temperature 3.85 1.01 9.77 (0.28)
Low temperature 3.74 1.06 10.69 (0.22)
How windy it will be 3.28 1.08 7.60 (0.47)
How cloudy it will be 2.74 1.08 14.38 (0.07)
Air pressure 2.21 1.13 10.81 (0.21)
SourcesSources
SourcesSources(regression analysis)(regression analysis)
Local TV Newscasts (Q3A)
Parameter Est. (SE)
Intercept 3.445***
Gender 0.237***
Age (years) 0.018***
Income (1,000s) 0.001
Education (years) -0.043
Percent time work outdoors 0.002
Percent time leisure outdoors 0.008***
AdjRSq 0.088
ProbF 0.000 *, **, *** significant at 10%, 5%, and 1% respectively
UseUse
UseUseOutdoor v. IndoorOutdoor v. Indoor
Planning
< 50% of Leisure Time Outdoors
(n = 187) Mean (SD)
50% + Leisure Time Outdoors (n = 194)
Mean (SD) Wilcoxon Z
Dress for the day 3.86 4.07 1.77 *
Planning for the weekend 3.41 4.09 5.98 ***
Vacation or travel 3.31 3.89 4.84 ***
Social activities 2.97 3.43 3.97 ***
House or yardwork 2.94 3.26 2.50 **
How to get to work/school/store
2.73 3.02 2.08 **
Job or business 2.46 3.06 4.19 *** 1 = “never”; 2 = “rarely”; 3 = “half the time”; 4 = “often”; 5 = “most of the time.”
*, **, *** significant at 10%, 5%, and 1% respectively
Adequacy of Current Levels Adequacy of Current Levels of Forecast Attributesof Forecast Attributes
Attribute Mean SD
Adequacy of updates 4 times a day 3.30 0.68
Adequacy of weather forecasts 5 days in advance
2.89 0.84
Adequacy of 80% correctness of one-day forecasts
2.88 0.81
Adequacy of geography detail to 30 miles by 30 miles
2.74 0.88
Stated Choice:Stated Choice:Attributes and Attribute Attributes and Attribute
LevelsLevels
•Dollars per year per household of $3, $8, $15, $24
•Budget constraint reminder
•20 versions of survey
•9 Stated Choice and 1 Stated Value question
Frequency One-Day Multiday Accuracy
Attribute Improvement
Level
Frequency of Updates (times
per day)
Accuracy of One-Day
Forecasts
Accuracy of Multiday
Forecasts Geographic
Detail
Baseline 4 80% 5 days 30 miles
Minimal 6 85% 7 days 15 miles
Medium 9 90% 10 days 7 miles
Maximum 12 95% 14 days 3 miles
StateStated d
ChoicChoicee
QuestQuest--
ionion
A-B Probit ModelA-B Probit Model
A B
Y-C Fr , One , Multi , Geog
Y-C Fr , One , Multi , Geog
Random Utility Model: x
Choose A if utility from U U
x x
1 1 1 2 2
2 1 2 1
,
,
A A A A A A
B B B B B B
k k k
A A B B
ij ij ij ij ij
ij ij ij ij
U
U
U
P P x x
P x x
univariate standard normal dist.function
x x
2 1
1 2
1 1
2
, 1, . . . , , , ij
ij ij
n Jk
ij ij ij iji j
x x
L k i J , P
Stated Choice QuestionStated Choice Question
A-B-Status Quo Model A-B-Status Quo Model (Conditional Probit)(Conditional Probit)
Random Utility Model (RUM) x
Example-Choose A over B and then stay with A over Status Quo
x x x x
x x x x
0
0 0
0 2 22 0
,
( ), ( )
2 , ;
k k
A B A
B A B A A A
B A A
U
P U U U U
P
Stated Value: Valuation Stated Value: Valuation QuestionQuestion
Stated Value (WTP) ModelStated Value (WTP) Model
Fr , One , Multi , Geog
x
Y-WTP Fr , One , Multi , Geog Y Fr , One , Multi , Geog
let
* * * *
* * *
0 * * * * * 0 0 0 0
* * 0 * 0
2 2* * 0 0
2
22 0
, ,
1( ) . . . ( )
~ . . . ,
fry
fr
y y
WTP f
U
U
WTP Fr Fr
WTP N Fr Fr
2
2y
Model EstimatesModel Estimates(t-ratios in parentheses)(t-ratios in parentheses)
ABO Biv.
Probit WTP Tobit Combined Frequency -0.049
(-10.0) 0.199 (0.8)
-0.067 (-16.4)
One Day 0.062 (16.4)
0.572 (4.0)
0.041 (13.3)
Multi-day 0.031 (6.5)
0.284 (1.2)
0.004 (1.1)
Geographic -0.007 (-4.8)
-0.272 (-4.0)
-0.031 (-25.6)
Cost -0.092 (-17.7)
-0.083 (-22.5)
Est. WTP (Est. std. err.)
$15.27 ($1.05)
$18.49 ($2.08)
$17.88 ($0.96)
N 3429 381 381
National Valuation National Valuation EstimateEstimate
Estimated household WTP $17.88
Number of Households 104,705,000
National WTP $1.872 B
Next StepsNext Steps THORPEX GrantTHORPEX Grant
Re-defining attribute sets and levelsRe-defining attribute sets and levels Temperature: 0-2 days 3-6 days 7-14 daysTemperature: 0-2 days 3-6 days 7-14 days Precipitation : 0-2 days 3-6 days 7-14 daysPrecipitation : 0-2 days 3-6 days 7-14 days Geographic SpecificityGeographic Specificity
National sample - ~1400 completesNational sample - ~1400 completes Internet based implementationInternet based implementation Probablistic forecast InformationProbablistic forecast Information Modeling and analysisModeling and analysis
non-linear in attribute levelsnon-linear in attribute levels random parameters random parameters socio-demographic characteristicssocio-demographic characteristics
Any Any Questions?Questions?
0101
THE REST OF THE SLIDES THE REST OF THE SLIDES HERE ARE EXTRAHERE ARE EXTRA
WE WON’T MAKE YOU WE WON’T MAKE YOU SUFFER THROUGH THE SUFFER THROUGH THE
REST REST (RIGHT NOW AT LEAST)(RIGHT NOW AT LEAST)
0101
What is Value?What is Value?
Neoclassical versus other approaches?Neoclassical versus other approaches?
Economic values and societal impacts, Economic values and societal impacts, e.g.,e.g.,
• lives savedlives saved• time savedtime saved• environmental valuesenvironmental values• impact on vulnerable populationsimpact on vulnerable populations
0101
What is Value?What is Value? Economic agentsEconomic agents
ConsumersConsumers ProducersProducers GovernmentGovernment
AssumptionsAssumptions1.1. People have rational preferencesPeople have rational preferences2.2. Individuals maximize utilityIndividuals maximize utility3.3. Firms maximize profitsFirms maximize profits4.4. Agents act independently using full Agents act independently using full
information Neoclassical theory includes or extends toNeoclassical theory includes or extends to
Competitive equilibriumCompetitive equilibrium Non-market and intrinsic valuesNon-market and intrinsic values Social welfare theory (incl. benefit-cost Social welfare theory (incl. benefit-cost
analysis)analysis) Value of information (VOI)Value of information (VOI)
What is Value?What is Value?
P$
Q
S
D
P*
Q*
Producers maximizing profits
by offering quantities for sale at different prices.
Consumers maximizing utility by buying
quantities at different prices.Equilibrium price (P*)
and quantity (Q*) determined by
interaction of Supply and Demand
What is Value?What is Value? PRODUCER PERSPECTIVE
P$
Q
S
D
P*
Q*
Producer Surplus =
Total Revenues
minus
Total Marginal Costs
PS
Total revenues = price x quantity
Total costs = sum of marginal
costs
What is Value?What is Value? CONSUMER PERSPECTIVE
P$
Q
S
D
P*
Q*
Consumer Surplus =
Total Benefits
minus
Total Expenditures
CS
Total benefits = sum of marginal
benefits
Total expenditures = price x quantity
What is Value?What is Value?
P$
Q
S
D
P*
Q*
CS
PS
Total social benefit = CS + PS
What is Value?What is Value?
P$
Q
S0
D0
P*
Q*
CS
PS D
1
S1
Q1
CS
PS
What is Value?What is Value? Neo-classical economics – utility theoryNeo-classical economics – utility theory Willingness to pay - WTPWillingness to pay - WTP
1
2 1
1 1 1, ; , ;
where implies "preferred"
2-WTP wV Y P w U V Y P
w w
11 1 , ;U V Y P w
1
1 2
11 2 1 1 2 2
,, ; s.t.
X XU U X X w P X P X Y
Q
S0
D0
P*
Q*
CS
PSD
1
S1
Q1
CS
11 1 1, ; , ; 2-WTP wV Y P w U V Y P
WTP: How much income could
be taken away from the
individual who receives
improved weather
forecasts while keeping him at the same level
of utility
What is Value?What is Value?
Q
S0
D0
P*
Q*
CS
PSD
1
S1
Q1
PS
What is Value?What is Value?
11 1 1, ; , ; 2-WT wP P+ SV Y P w U V Y P
PPS =WT
How Are Values Measured?How Are Values Measured?Murphy, A.H., 1994: Assessing The Economic Murphy, A.H., 1994: Assessing The Economic
Value Of Weather Forecasts: An Overview Of Value Of Weather Forecasts: An Overview Of Methods, Results And Issues. Methods, Results And Issues. Meteorological Applications, 1(2), 69-73.Meteorological Applications, 1(2), 69-73.
Anaman, K.A., D.J. Thampapillai, A. Henderson-Anaman, K.A., D.J. Thampapillai, A. Henderson-Sellers, P.F. Noar, And P.J. Sullivan, 1995: Sellers, P.F. Noar, And P.J. Sullivan, 1995: Methods For Assessing The Benefits Of Methods For Assessing The Benefits Of Meteorological Services In Australia. Meteorological Services In Australia. Meteorological Applications, 2(1), 17-29 Meteorological Applications, 2(1), 17-29
Macauley, M.K., 1997: Some Dimensions Of Macauley, M.K., 1997: Some Dimensions Of The Value Of Weather Information: General The Value Of Weather Information: General Principles And A Taxonomy Of Empirical Principles And A Taxonomy Of Empirical Approaches. Report Of Workshop On The Approaches. Report Of Workshop On The Social And Economic Impacts Of Weather, Social And Economic Impacts Of Weather, Boulder, Co.Boulder, Co.
How Are Values Measured?How Are Values Measured? Revealed preferenceRevealed preference
market pricesmarket prices hedonic methodshedonic methods
Stated preferenceStated preference stated value (contingent valuation)stated value (contingent valuation) stated choicestated choice
Prescriptive StudiesPrescriptive Studies
Descriptive StudiesDescriptive Studies