Longitudinal Data Analysis of Age and Happiness Ning Li Australian - PowerPoint PPT Presentation

Longitudinal Data Analysis of Age and Happiness Ning Li Australian Mathematical Sciences Institute Oceania Stata Workshop/Conference 20 Aug 2019, Parramatta Sydney

Age & Happiness • Happiness is a good indicator of the quality of life • Is Happiness Related to Age ? How? • Many Studies • No Agreed Opinion https://encrypted-tbn0.gstatic.com/images?q=tbn:ANd9GcTl_QEwWoQfZR_rMwesX-cT0dY5vTp1kTf_8-C7crp_GikwbR2 1/14

Mixed Results • U (Blanchflower & Oswald 2004 p1359, 2005 p311, 2008 p1733, 2009 p486, 2011 p9, …) • Constant (Costa et al. 1987 p54, Easterlin & Schaeffer 1999, Myers 2000 Fig5, Carstensen et al 2000 p651, …) • Increasing (Mroczek & Kolarz 1998 p1346, Lacey et al 2006 p167, …) • Inverted U (Mroczek & Spiro 2005 p154, Chen 2001 p57, …) • Wave shape (Horley & Lavery 1994 p277, Frijiters & Beatton 2012, …) • U in pooled OLS, decline in FE regression (Ferrer-i-Carbonell & Frijters 2004 p655, Landeghem 2008 p1, Kassenboehmer & Haisken-DeNew 2012 p235, Frijters & Beatton 2012 p540, Wood & Li 2013, … ) • U in pooled OLS, U in FE OLS (Clark 2007 p11, Kristoffersen 2013 p16, …) 2/14

Reconciling U , -- , ∩ , W Shapes 90 85 Inverted U 80 Increase Decrease 75 Constant 70 20 40 60 80 100 Age 3/14

“The U-bend of life”. Economist [London, England]18 Dec. 2010: 33+. The Economist Historical Archive , 1843-2013. Web. 19 Aug.2017 4/14

The U-shape Debate “Is well-being U-shaped over the life cycle?” Blanchflower & Oswald (2008) Social Science & Medicine 66 , 1733-1749 Yes. SWB is U-shaped over the Life Cycle. N. Glenn (2009) Social Science & Medicine 69 , 481-485 No. The U-shape is a Result of the Use of Inappropriate Control Variables. Blanchflower & Oswald (2009) Social Science & Medicine 69 , 486–488 The U-Shape Remains without any Controls. 8/14

id H m21 m23 m22 m24 m25 m26 m27 m28 m29 1 0 u u u u u u u u u Glenn’s Example ( SSM 69 (2009), p482) 2 1 u u u u u u u u m 3 2 u u u u u u u m m 4 3 u u u u u u m m m Happiness = α + β Age 5 4 u u u u u m m m m 6 5 u u u u m m m m m 7 6 u u u m m m m m m 8 7 u m u m m m m m m 9 8 u m m m m m m m m 10 9 m m m m m m m m m 11 0 u u u u u u u u u 12 1 u u u u u u u u m 13 2 u u u u u u u m m 14 3 u u u u u u m m m 15 4 u u u u u m m m m 16 5 u u u u m m m m m 17 6 u u u m m m m m m 18 7 u m u m m m m m m Happiness = α + β Age 19 8 u m m m m m m m m 20 9 m m + ᵧ (Marital Status) m m m m m m m 21 0 u u u u u u u u u 22 1 u u u u u u u u m 23 2 u u u u u u u m m . . 97 6 u u u m m m m m m 98 7 u m u m m m m m m 99 8 u m m m m m m m m 100 9 m m m m m m m m m 6/14

Glenn’s Example ( SSM 69 (2009), p482 ) Happiness = α + β Age Average Happiness Observed Average Happiness 21 22 26 28 Estimated Average Happiness 23 24 25 27 29 Age Happiness = α + β Age + ᵧ (Marital Status) 7/14

The U-shape Debate “Is well-being U-shaped over the life cycle?” Blanchflower & Oswald (2008) Social Science & Medicine 66 , 1733-1749 Yes. SWB is U-shaped over the Life Cycle. N. Glenn (2009) Social Science & Medicine 69 , 481-485 No. The U-shape is a Result of the Use of Inappropriate Control Variables. Blanchflower & Oswald (2009) Social Science & Medicine 69 , 486–488 The U-Shape Remains without any Controls. 8/14

The Age-Happiness Mystery (Happiness) i Regression Curves With & Without Controlling for Individual Effects = α + β (Age) ἰ + ε ἰ 90 OLS Using W1 Data 85 80 75 70 FE regression using W1-W11 Data 65 (Happiness) it 60 = α ἰ + β (Age) ἰt + ε ἰt 20 40 60 80 100 Age 9/14

Some Explanations in the Literature about the Mystery “ The data does not bear any “ T h e f o u n d e f f e c t o f a g e i n f i x e d - useful information to support one e f f e c t r e g r e s s i o n s i s s i m p l y t o o conclusion but not the other.” l a r g e a n d t o o o u t o f l i n e w i t h e v e r y t h i n g (Ree&Alessie 2011, p182) e l s e w e k n o w t o b e b e l i e v a b l e . ” “…the negative relationship ( F r i j t e r s & B e a t t o n 2 0 0 8 , p 2 2 ) between happiness and age is a “ T h e o t h e r w i s e misconception ” (Lacey et al. 2012, p647) s e e m i n g l y r o b u s t i a n g p e o o U l - e s h d a O p L e S e r f e f g e r c t e s o s i n o l n i f s e i s s a r t e i s f f u a t e c d t i o … n e w f h f e c e t n s c . ” (Kassenboehmera, S. C., & Haisken-DeNew 2012) o n t r o l l i n g f o r p a n e l f i x e d ” (Frijters & Beatton 2008, p18) a t a d e h t n i g n o w r g n h i t e m o S “ 10/14

A Statistical Explanation • The U shape, found in the cross-section data, represents the average happiness of various groups of people at different stages of their lives. • Happiness varies more between people than within people over time, hence the OLS result in pooled data resembles that in cross-section data. • The decline pattern, found in the FE regression, represents the average happiness of the same group of people in their life course when they age. • The yearly change from the fixed effects regression of happiness on age = or ≈ the weighted average of yearly change in the observed data. 11/14

A Graph With An Ambiguous Axis (Happiness) i Regression Curves With & Without Controlling for Individual Effects = α + β (Age) ἰ + ε ἰ 90 OLS Using W1 Data 85 80 75 70 FE regression using W1-W11 Data 65 (Happiness) it 60 = α ἰ + β (Age) ἰt + ε ἰt 20 40 60 80 100 Age 12/14

From (Age, Happiness) To (Cohort, Ageing, Happiness) g n i g e n A g i e g A 13/14