Measuring well-being for public policy: preferences or experiences

Measuring Happiness and Making Policy
Professor Paul Dolan
Tanaka Business School
Imperial College London
Income versus happiness
Well-being measures, standardized scores
0.5
0.4
0.3
Well-being measures, standardized scores
0.5
0.4
A
0.2
0.1
0.1
0
0
Income
-0.1
Life Sat
-0.1
-0.2
-0.2
-0.3
-0.3
-0.4
-0.4
-0.5
-0.5
Well-being measures, standardized scores
0.5
0.4
0.4
C
0.2
0.1
0.1
0
Over 70
Has children
D
0
Income
Life Sat
-0.1
-0.2
-0.2
-0.3
-0.3
-0.4
-0.4
-0.5
-0.5
Commute >
1hr/day
Life Sat
0.3
0.2
-0.1
Income
Well-being measures, standardized scores
0.5
0.3
B
0.3
0.2
Income
Life Sat
Has degree
How do we decide?



Is the measure conceptually appropriate?
Is the measure normatively relevant?
Is the measure empirically useful?
Is the measure conceptually appropriate?

Distinguish between good life and good for the individual
Prudential value is suitable for most policy applications
– Both measures mostly contain prudential concerns

The measure should be a complete account of well-being

– There is more to life than income (and happiness?)

The measure should measure what it purports to
– Allows us to make intra and inter-personal comparisons

There is no gold standard but there is ‘validity’
– Both measures converge with and predicts health etc.
Is the measure normatively relevant?

The measure must be acceptable to policy makers and
to the general public, at least in time

Income is attractive because the individual judges what
will be in his best interests
But may be questioned if it does not lead to LS
And if preferences are misguided, myopic etc.




Support for LS may depend on language used
LS may be better than happiness, and misery better still?
Is the measure empirically useful?

Is it intra- and interpersonally comparable?
– Both shaped by expectations, shaped by circumstances
 Brain imaging and physiological measures may help LS

Is it cardinal?
– LS performs better than income here

Is measurement error low?
– Both measures reasonably stable for most sub-groups

Is it sensitive?
– Income changes may not be meaningful but global LS may not
pick up small effects of policy

Is it practical?
– Both relatively easy to collect
Conclusion

Using different measures may result in different policies


No single measure is ever likely to satisfy all criteria
But well-being measures are used in policy as if they do

Choice really depends on which criteria matter most
– Some criteria may be more important in some contexts
– And there may be various trade-offs between the criteria


We need more conceptual clarity and empirical evidence
But subjective evaluations may well be the way forward