The Origins of Happiness

Notes to Pages 46–54

27. See also Clark and D’Ambrosio (2015). In many ways experi-
    mental data involve fewer problems than naturalistic data. In this case
    the work is not based on happiness regressions, but rather stated prefer-
    ences over hypothetical scenarios involving income distributions that
    an imaginary grandchild will face (in Johannsson- Stenman, Carlsson,
    and Daruvala [2002]) or leaky- bucket experiments where individuals
    are asked to indicate the amount of “lost money” that they are willing
    to accept for a transfer of money from a richer to a poorer individ-
    ual (see for example Amiel, Creedy, and Hurn [1999]). The conclusion
    from this work is that individuals do seem to have preferences over in-
    come inequality, and not only because their own income or their rela-
    tive income is affected. However it does seem to be difficult to quantify
    exactly how much this income inequality matters.


28. See full results in online Table A2.3.


29. If we add highest qualification, the R^2 of the equation rises
    from 0.26 to 0.31; see online Table A2.3.


30. If H = αlogY where H is happiness and Y income, dH/dY = α/ Y.


31. See also Layard (2006).

Chapter 3. Education

1. For earlier work on this issue, see online Annex 3a. On the issue
   of credentialism, note that measured IQ has risen sharply over time
   (Pietschnig and Voracek [2015]).


2. On the United States see Oreopoulos and Petronijevic (2013).
   On the UK, see Blundell, Green, and Jin (2016) and Walker and Zhu
   (2008).


3. It may also lead to more enjoyable jobs (which are therefore less
   well paid). The surveys provide no data on this.


4. No qualifications, Level 1 (CSE and O- level equivalent [grades
   (D– G)]), Level 2 (O- level equivalent [grades A*– C]), Level 3 (A- level
   equivalent), and degree or above.


5. We first run the following equation:

Log Y = α + ^5 j = 1 βjEducj + etc.

where Y is income and Educj are education dummies for each

level of qualification. We then use the coefficients on each edu-

cation dummy to create a simple continuous education variable.

6. This is the standard deviation of years of schooling in the BHPS.


7. See also Oreopoulos and Salvanes (2011).