Abstract
De Finetti suggested that scoring rules – namely, loss functions by which a forecaster is virtually charged depending on the degree of inaccuracy of his predictions – could be employed also to provide a compelling argument for probabilism. However, De Finetti’s choice of a specific scoring rule for this purpose (Brier’s quadratic rule) appears somewhat arbitrary, and the general pragmatic flavour of the argument – which makes it a variant of the well-known “Dutch Book Theorem” – has been deemed unsuitable for an epistemic justification of probabilism. In this paper we suggest how Brier’s rule may be justified on epistemic grounds by means of a strategy that is different from the one usually adopted for this purpose (e.g., in Joyce 1998), taking advantage of a recent characterization result concerning distance functions between real-valued vectors (D’Agostino and Dardanoni 2008).
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
Joyce (1998), p. 584.
- 2.
As Murphy and Winkler put it: “Categorical or deterministic forecasts suffer from two serious deficiencies. First, a forecaster or forecasting system is seldom, if ever, certain of which event will occur. Second, categorical forecasts do not provide users of the forecasts with the information that they need to make rational decisions in uncertain situations.”(Murphy and Winkler 1984, p. 489).
- 3.
Our use of the expression “distance function”, in this paper, is rather loose and refers to any continuous function of two real-valued vectors (of the same finite size) which can be intuitively regarded as measuring their distance, including functions which do not fully satisfy the standard textbook definition. Such non-standard “distance” functions have often been used to solve measuring problems in several applications areas.
- 4.
Some authors prefer to define scoring rules so as the forecaster’s goal is that of maximizing his score. Trivial algebraic manipulation is sufficient to turn one scenery into the other.
- 5.
For this purpose, it is sufficient to add \((1 - (x + y))/2\) to both x and y.
- 6.
- 7.
On this point, see the already cited (Joyce 1998).
- 8.
- 9.
Observe that any scoring rule based on the generalized α-power (with α > 1) of the absolute differences between subjective estimates and observed outcomes is proper. On this point see (Selten 1998).
References
Atkinson A (1983) The measurement of economic mobility. In: Atkinson A (ed) Social justice and public policy. Wheatsheaf Books Ltd., London
Brier G (1950) Verfication of forecasts expressed in terms of probability. Mon Wea Rev 78:1–3
D’Agostino M, Dardanoni V (2009) What’s so special about Euclidean distance? A characterization with applications to mobility and spatial voting. Soc Choice Welfare 33:211–233
Dardanoni V (1993) Measuring social mobility. J Econ Theory 61:372–394
Debreu G (1960) Topological methods in cardinal utility. In: Arrow SK, Suppes P (eds) Mathematical methods in the social sciences. Stanford University Press, Stanford, CA
Epstein L, Tanny S (1980) Increasing generalized correlation: a definition and some economic consequences. Can J Econ 13:16–34
de Finetti B (1974) Theory of probability, vol 1. Wiley, New York
Joyce JM (1998) A non-pragmatic vindication of probabilism. Philos Sci 65(4):575–603
Kennedy R, Chihara C (1979) The dutch book argument: its logical flaws, its subjective sources. Philos Stud 36:19–33
Lindley D (1982) Scoring rules and the inevitability of probability. Int Stat Rev 50:1–26
Murphy A, Winkler R (1984) Probability forecasting in meteorology. J Am Stat Assoc 79:489–500
Offerman T, Sonnemans J, Kuilen Gv, Walker P (2009) A truth-serum for non-Bayesians: correcting proper scoring rules for risk attitudes. Resource document. Rev. Econ. Stud., forthcoming
Rosenkrantz R (1981) Foundations and applications of inductive probability. Ridgeview, Atascadero, CA
Savage L (1971) Elicitation of personal probabilities and expectations. J Am Stat Assoc 66: 783–800
Selten R (1998) Axiomatic characterization of the quadratic scoringrule. Experiment Econ 1:43–62
Tchen A (1980) Inequalities for distributions with given marginals. Ann Probability 8:814–827
Tverski A, Krantz DH (1970) The dimensional representation and the metric structure of similarity data. J Math Psychol 7:572–596
Winkler R (1969) Scoring rules and the evaluation of probability assessors. J Am Stat Assoc 64(327):1073–1078
Acknowledgements
We wish to thank Wolfgang Spohn for very useful suggestions.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2009 Springer Science+Business Media B.V.
About this chapter
Cite this chapter
D’Agostino, M., Sinigaglia, C. (2009). Epistemic Accuracy and Subjective Probability. In: Suárez, M., Dorato, M., Rédei, M. (eds) EPSA Epistemology and Methodology of Science. Springer, Dordrecht. https://doi.org/10.1007/978-90-481-3263-8_8
Download citation
DOI: https://doi.org/10.1007/978-90-481-3263-8_8
Published:
Publisher Name: Springer, Dordrecht
Print ISBN: 978-90-481-3262-1
Online ISBN: 978-90-481-3263-8
eBook Packages: Humanities, Social Sciences and LawPhilosophy and Religion (R0)