Forecasting and the importance of being uncertain 1

Forecasting and theimportance of being

Rob J Hyndman

uncertain

Forecasting and the importance of being uncertain 2

Outline

1 Dangers and difficulties of forecasting

2 Forecasting mortality and fertility

3 Forecasting the PBS

4 Forecasting peak electricity demand

5 Final thoughts

Forecasting and the importance of being uncertain Dangers and difficulties of forecasting 3

Outline

1 Dangers and difficulties of forecasting

2 Forecasting mortality and fertility

3 Forecasting the PBS

4 Forecasting peak electricity demand

5 Final thoughts

Forecasting and the importance of being uncertain Dangers and difficulties of forecasting 4

Forecasters are to blame!

News report on 16 August 2006A Russian woman is suing weather forecastersfor wrecking her holiday. A court in Uljanovskheard that Alyona Gabitova had been promised28 degrees and sunshine when she planned acamping trip to a local nature reserve,newspaper Nowyje Iswestija said.But it did nothing but pour with rain the wholetime, leaving her with a cold. Gabitova hasasked the court to order the weather service topay the cost of her travel.

Forecasting and the importance of being uncertain Dangers and difficulties of forecasting 5

Those “unforeseen events”

Precautions should be taken against

running into unforeseen occurrences or

events. (Horoscope, New York Times)

We are ready for any unforeseen event

which may or may not occur.

(Dan Quayle)

Forecasting and the importance of being uncertain Dangers and difficulties of forecasting 5

Those “unforeseen events”

Precautions should be taken against

running into unforeseen occurrences or

events. (Horoscope, New York Times)

We are ready for any unforeseen event

which may or may not occur.

(Dan Quayle)

Forecasting and the importance of being uncertain Forecasting mortality and fertility 6

Outline

1 Dangers and difficulties of forecasting

2 Forecasting mortality and fertility

3 Forecasting the PBS

4 Forecasting peak electricity demand

5 Final thoughts

Forecasting and the importance of being uncertain Forecasting mortality and fertility 7

Mortality: female

0 20 40 60 80 100

−10

−8

−6

−4

−2

Australia: female death rates (1950−2003)

Age

Log

deat

h ra

te

Forecasting and the importance of being uncertain Forecasting mortality and fertility 7

Mortality: female

0 20 40 60 80 100

−10

−8

−6

−4

−2

Australia: forecast female log death rates (2004, 2023)

Age

Log

deat

h ra

te

Forecasting and the importance of being uncertain Forecasting mortality and fertility 7

Mortality: female

0 20 40 60 80 100

−10

−8

−6

−4

−2

Australia: forecast female log death rates (2004, 2023)

Age

Log

deat

h ra

te

●

●

●●●

●

●

●●●

●

●●●

●●

●●●●●●●●

●●●●●●

●●●

●●●●●●●

●●●●●●●●●●●●●●

●●●

●●●●●●●

●●●●●

●●●●●●●●●●●●●

●●●

●●●

●●●●●●●●●

●●●●

Forecasting and the importance of being uncertain Forecasting mortality and fertility 8

Fertility

15 20 25 30 35 40 45 50

050

100

150

200

250

Australia fertility rates (1921−2003)

Age

Fer

tility

rat

e

Forecasting and the importance of being uncertain Forecasting mortality and fertility 8

Fertility

15 20 25 30 35 40 45 50

050

100

150

200

250

Australia: forecast fertility rates (2004, 2023)

Age

Fer

tility

rat

e

Forecasting and the importance of being uncertain Forecasting mortality and fertility 8

Fertility

15 20 25 30 35 40 45 50

050

100

150

200

250

Australia: forecast fertility rates (2004, 2023)

Age

Fer

tility

rat

e

●●

●●

●●

●●

●

●

●

●●

●● ● ●

●●

●

●

●

●

●

●

●●

●● ● ● ● ● ● ●

Forecasting and the importance of being uncertain Forecasting mortality and fertility 9

Forecasting mortality and fertility

Key papers

Erbas, B., Hyndman, R.J., and Gertig, D.M. (2007)“Forecasting age-specific breast cancer mortalityusing functional data models”. Statistics inMedicine, 26(2), 458-470.

Hyndman, R.J., and Ullah, Md. S. (2007) “Robustforecasting of mortality and fertility rates: afunctional data approach”. ComputationalStatistics and Data Analysis, to appear.

Since 2005, these methods have formed the basis of allofficial cancer mortality and incidence forecasts fromthe Australian Institute of Health and Welfare.

Forecasting and the importance of being uncertain Forecasting mortality and fertility 9

Forecasting mortality and fertility

Key papers

Erbas, B., Hyndman, R.J., and Gertig, D.M. (2007)“Forecasting age-specific breast cancer mortalityusing functional data models”. Statistics inMedicine, 26(2), 458-470.

Hyndman, R.J., and Ullah, Md. S. (2007) “Robustforecasting of mortality and fertility rates: afunctional data approach”. ComputationalStatistics and Data Analysis, to appear.

Since 2005, these methods have formed the basis of allofficial cancer mortality and incidence forecasts fromthe Australian Institute of Health and Welfare.

Forecasting and the importance of being uncertain Forecasting mortality and fertility 10

Forecasting population

● ● ●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●

1980 1990 2000 2010 2020

78

910

1112

13

Total females

Year

Pop

ulat

ion

(mill

ions

)

Forecasting and the importance of being uncertain Forecasting mortality and fertility 10

Forecasting population

● ● ● ● ●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●●

●

1980 1990 2000 2010 2020

78

910

1112

13

Total males

Year

Pop

ulat

ion

(mill

ions

)

Forecasting and the importance of being uncertain Forecasting the PBS 11

Outline

1 Dangers and difficulties of forecasting

2 Forecasting mortality and fertility

3 Forecasting the PBS

4 Forecasting peak electricity demand

5 Final thoughts

Forecasting and the importance of being uncertain Forecasting the PBS 12

Forecasting the PBS

Estimation of forward estimates for thePharmaceutical Benefit SchemeDepartment of Health and Ageing

$5 billion budget. Underforecasted by$500–$800 million in 2000 and 2001.

Thousands of products. Seasonal demand.

Subject to covert marketing, volatileproducts, uncontrollable expenditure.

All forecasts being done with the FORECASTfunction in MS-Excel applied to 10 year olddata!

Forecasting and the importance of being uncertain Forecasting the PBS 12

Forecasting the PBS

Estimation of forward estimates for thePharmaceutical Benefit SchemeDepartment of Health and Ageing

$5 billion budget. Underforecasted by$500–$800 million in 2000 and 2001.

Thousands of products. Seasonal demand.

Subject to covert marketing, volatileproducts, uncontrollable expenditure.

All forecasts being done with the FORECASTfunction in MS-Excel applied to 10 year olddata!

Forecasting and the importance of being uncertain Forecasting the PBS 12

Forecasting the PBS

Estimation of forward estimates for thePharmaceutical Benefit SchemeDepartment of Health and Ageing

$5 billion budget. Underforecasted by$500–$800 million in 2000 and 2001.

Thousands of products. Seasonal demand.

Subject to covert marketing, volatileproducts, uncontrollable expenditure.

All forecasts being done with the FORECASTfunction in MS-Excel applied to 10 year olddata!

Forecasting and the importance of being uncertain Forecasting the PBS 12

Forecasting the PBS

Estimation of forward estimates for thePharmaceutical Benefit SchemeDepartment of Health and Ageing

$5 billion budget. Underforecasted by$500–$800 million in 2000 and 2001.

Thousands of products. Seasonal demand.

Subject to covert marketing, volatileproducts, uncontrollable expenditure.

All forecasts being done with the FORECASTfunction in MS-Excel applied to 10 year olddata!

Forecasting and the importance of being uncertain Forecasting the PBS 13

Forecasting the PBS

Estimation of forward estimates for thePharmaceutical Benefit SchemeDepartment of Health and Ageing

We developed some new time seriesmodels — automated exponentialsmoothing state space modelling appliedto about 100 product groups.

Models allowed for time-changing trendand seasonal patterns and provideprediction intervals.

Forecast error now a few $million per year.

Forecasting and the importance of being uncertain Forecasting the PBS 13

Forecasting the PBS

Estimation of forward estimates for thePharmaceutical Benefit SchemeDepartment of Health and Ageing

We developed some new time seriesmodels — automated exponentialsmoothing state space modelling appliedto about 100 product groups.

Models allowed for time-changing trendand seasonal patterns and provideprediction intervals.

Forecast error now a few $million per year.

Forecasting and the importance of being uncertain Forecasting the PBS 13

Forecasting the PBS

Estimation of forward estimates for thePharmaceutical Benefit SchemeDepartment of Health and Ageing

We developed some new time seriesmodels — automated exponentialsmoothing state space modelling appliedto about 100 product groups.

Models allowed for time-changing trendand seasonal patterns and provideprediction intervals.

Forecast error now a few $million per year.

Forecasting and the importance of being uncertain Forecasting the PBS 14

Forecasting the PBSTotal monthly scripts: concession copayments

Cardiovascular system drugs

Year

x 10

00

1995 2000 2005

2040

6080

100

120

140

80% prediction intervals

Forecasting and the importance of being uncertain Forecasting the PBS 15

Forecasting the PBSKey papers

Hyndman, R.J., Koehler, A.B., Snyder, R.D., andGrose, S. (2002) “A state space framework forautomatic forecasting using exponential smoothingmethods”. International J. Forecasting, 18(3),439–454.Hyndman, R.J., Koehler, A.B., Ord, J.K., and Snyder,R.D. (2005) “Prediction intervals for exponentialsmoothing using two new classes of state spacemodels”. Journal of Forecasting, 24(1), 17–37.

Forthcoming monographForecasting with exponential smoothing: the state spaceapproach by Hyndman, Koehler, Ord and Snyder(Springer, 2008).

Forecasting and the importance of being uncertain Forecasting the PBS 15

Forecasting the PBSKey papers

Hyndman, R.J., Koehler, A.B., Snyder, R.D., andGrose, S. (2002) “A state space framework forautomatic forecasting using exponential smoothingmethods”. International J. Forecasting, 18(3),439–454.Hyndman, R.J., Koehler, A.B., Ord, J.K., and Snyder,R.D. (2005) “Prediction intervals for exponentialsmoothing using two new classes of state spacemodels”. Journal of Forecasting, 24(1), 17–37.

Forthcoming monographForecasting with exponential smoothing: the state spaceapproach by Hyndman, Koehler, Ord and Snyder(Springer, 2008).

Forecasting and the importance of being uncertain Forecasting peak electricity demand 16

Outline

1 Dangers and difficulties of forecasting

2 Forecasting mortality and fertility

3 Forecasting the PBS

4 Forecasting peak electricity demand

5 Final thoughts

Forecasting and the importance of being uncertain Forecasting peak electricity demand 17

Peak electricity demand

Forecasting and the importance of being uncertain Forecasting peak electricity demand 17

Peak electricity demandSA demand (first 3 weeks of January 2007)

Date in January

SA

dem

and

(GW

)

1.5

2.0

2.5

1 2 3 4 5 6 7 8 9 10 12 14 16 18 2011 13 15 17 19 21

Forecasting and the importance of being uncertain Forecasting peak electricity demand 17

Peak electricity demand

2.8 2.9 3.0 3.1 3.2 3.3 3.4

0.00

00.

002

0.00

40.

006

0.00

8 Density of demand

Demand (GW)

Den

sity

● ● ●●● ●● ● ●● ●● ●● ●●● ●●● ●

Daily maxima

Forecasting and the importance of being uncertain Final thoughts 18

Outline

1 Dangers and difficulties of forecasting

2 Forecasting mortality and fertility

3 Forecasting the PBS

4 Forecasting peak electricity demand

5 Final thoughts

Forecasting and the importance of being uncertain Final thoughts 19

Final thoughts

Uncertainty statements are essentialwhen making predictions.

Good statistical models can often beapplied to diverse applications.Working in demography, oncology,epidemiology, pharmaco-epidemiology,energy, meteorology, and other fieldskeeps life interesting!

A good forecaster is not smarter than everyoneelse, he merely has his ignorance betterorganised.

(Anonymous)

Forecasting and the importance of being uncertain Final thoughts 19

Final thoughts

Uncertainty statements are essentialwhen making predictions.Good statistical models can often beapplied to diverse applications.

Working in demography, oncology,epidemiology, pharmaco-epidemiology,energy, meteorology, and other fieldskeeps life interesting!

A good forecaster is not smarter than everyoneelse, he merely has his ignorance betterorganised.

(Anonymous)

Forecasting and the importance of being uncertain Final thoughts 19

Final thoughts

Uncertainty statements are essentialwhen making predictions.Good statistical models can often beapplied to diverse applications.Working in demography, oncology,epidemiology, pharmaco-epidemiology,energy, meteorology, and other fieldskeeps life interesting!

A good forecaster is not smarter than everyoneelse, he merely has his ignorance betterorganised.

(Anonymous)

Forecasting and the importance of being uncertain Final thoughts 19

Final thoughts

Uncertainty statements are essentialwhen making predictions.Good statistical models can often beapplied to diverse applications.Working in demography, oncology,epidemiology, pharmaco-epidemiology,energy, meteorology, and other fieldskeeps life interesting!

A good forecaster is not smarter than everyoneelse, he merely has his ignorance betterorganised.

(Anonymous)

Forecasting and the importance of being uncertain Final thoughts 19

Final thoughts

Uncertainty statements are essentialwhen making predictions.Good statistical models can often beapplied to diverse applications.Working in demography, oncology,epidemiology, pharmaco-epidemiology,energy, meteorology, and other fieldskeeps life interesting!

A good forecaster is not smarter than everyoneelse, he merely has his ignorance betterorganised.

(Anonymous)