Seattle Public Utilities Water Demand Forecast Model

1
Seattle Public Utilities Water Demand Forecast
Model

• July 6, 2006
• Bruce Flory, SPU Principal Economist

2
Introduction

• Goals of Demand Forecasting
• Literature Review
• Model Structure, Inputs and Assumptions
• Modeling Uncertainty

3
Goals

• Good Decisions
• Identify causal factors
• Account for uncertainty
• Flexible
• Scenarios uncertainty
• Transparent
• Transferable

4
Literature Review

• Sources
• Web research
• Academic literature on demand forecasting, price,
  and conservation
• Other water electric utilities and consultants
• Forecasting methods were reviewed from
• Boston Chicago
  Columbus
• Denver East Bay MUD
  Eugene
• Los Angeles MWDSC
  New York
• Phoenix Portland
  San Diego
• San Francisco Tacoma
  Tampa Bay
• Tucson Washington DC

5
Lit Review Forecast Methods

• Trend Analysis
• Per Capita Flow Factors
• Sectoral Disaggregation
• Fixed Flow Factor Approach
• End Use Models
• Econometric Modeling
• Variable Flow Factor Approach

6
Lit Review - Conclusions

• Econometric modeling is the technique of choice
  given adequate resources (time, expertise, and
  data) and general agreement on the proper model
  specification.
• Such agreement does not exist
• Correct price variable
• Difficulty with statistically separating out
  impacts of price, code, programs, public
  information, new technology
• State of the art is unsettled
• However, much more agreement exists on the
  magnitude of price and income elasticities

7
Lit Review Conclusions (cont.)

• Take advantage of the econometric analysis done
  by others using a Variable Flow Factor approach
• Best achieves goals
• Assumptions can easily be made and varied (for
  sensitivity analysis)
• Simple yet takes account of many explanatory
  variables
• Modest data requirements

8
Lit Review Elasticities
9
Variable Flow Factor Model

• Current water demand flow factors by sector for
  Seattle and each wholesale customer.
• Impacts of variables such as price, income and
  conservation on water flow factors for each
  sector over time.
• Forecasts of households and employment.

10
Inputs for Variable Flow Factor Model

• For Base Year Water Flow Factors
• Current consumption
• Current households and employment
• Other Inputs Affecting Future Flow Factors
• Future income growth (forecast)
• Income elasticity of demand (literature review)
• Future growth in water prices (forecast)
• Price elasticity of demand (literature review)
• Future conservation savings (from CPA model)
• Conservation overlap function

11
Forecast Model Inputs (cont.)

• Forecasts of households and employment
• Source PSRC
• Forecasts extrapolated beyond 2030
• Other Adjustments to Forecast
• Forecast of non-revenue water
• Forecast of other sources of supply
• Potential new wholesale customers
• Block contracts

12
Demand Model Structure 1
13
Demand Model Structure 2
14
Demand Model Structure 3
15
Actual and Forecast Water Demand 1980-2060
16
Uncertainty

• High confidence in forecast, yet....
• Actual unlikely to be exactly as forecast
• Sources of uncertainty
• Forecasts of input variables
• Elasticities
• Other assumptions

17
Discrete vs. Continuous

• Discrete Specific events that produce
  significant sometimes abrupt changes
• Example Weyerhaeuser Water Right
• Handled by running individual what-if scenarios
  
• No probabilities assigned
• Continuous uncertainty that surrounds many inputs
  to the model
• Represented by probability distribution around a
  mean value

18
Continuous Uncertainty
19
Model Inputs with Continuous Uncertainty

• Projections of growth in single and multi-family
  households
• Projections of employment growth
• Projected annual growth rate in real water prices
• Price elasticities of demand
• Projected annual growth rate in real household
  income
• Income elasticity of demand
• Projected conservation savings

20
Modeling Demand Uncertainty

• Estimate range of uncertainty around model inputs
• Assign probability distributions
• Run Monte Carlo simulations
• 10,000 iterations

21
(No Transcript)
22
(No Transcript)
23
(No Transcript)
24
Implications
25
Seattle Public Utilities Water Demand Forecast
Model

• Discussion
• Q A