This is one of the big controversies in happiness economics.
I wonder if this effect (which directly contradicts Easterlin's findings that money has no effect on happiness past a certain point; here, they contend that there is no saturation point and that happiness evolves with the logarithm of income) could be somewhat due to the way they framed the question: "assume you are on a ladder of happiness with 10 steps, which one would you say you are on; then tell us how much you earn". It is widely known in behavioral economics that framing effects can have a huge effect on the way people respond ([1] [2], and so on), so I wouldn't be surprised than the way you ask people to evaluate their own happiness could explain at least part of the difference.
One of the conclusions of the Easterlin Paradox is that people care more about how much they earn compared to their neighbors than the actual amount. I feel like asking the question this way (imagining their happiness as a ladder with 10 steps, then thinking about their income) would directly lead to people implicitly internalizing this comparison in some kind of mental model where higher wealth = more happiness because they're trying to imagine what the best life possible could be: "ah I'm pretty happy right now, but if I had a twice as much money I think I would be happier though, so surely I can't be at the last step at the ladder. Actually, people who have twice as much than that should be even happier, so I'll add some steps on top and say I'm a 6 right now".
This sounds plausible to me because while the authors conclude there is no saturation point where income doesn't bring more happiness, this is a scale from 1 to 10 so surely some people must rate themselves a 10. What would those who earn twice as much as them think then? This solution could be that "being on top of the ladder" in people's mind is somehow conflated with "being on top of the income distribution", with all the other levels being computed relatively to that. In other terms, this framing may incite people to evaluate their happiness on a cross-sectional level, with Easterlin's Paradox being precisely that the income-happiness relationship exists only at the cross-sectional level but not at the longitudinal level.
The debate between the two models (Easterlin vs. Stevenson and Wolfers) has been a longstanding debate in happiness economics, with both sides having confirmed their findings with multiple data sets [3]. Maybe they're both right in a sense and the answer just depends on which definition of happiness you're asking people to evaluate themselves with (relative vs. on an absolute level)?
[1] http://economistsview.typepad.com/economistsview/2006/08/the...
[2] http://www.adsavvy.org/the-power-of-framing-effects-and-othe...
[3] http://en.wikipedia.org/wiki/Easterlin_paradox