[scarmig]

To figure out the causality, you'd need an experiment where people are randomly removed from relationships though random external events.

We do have an imperfect example of this, in the case of widowers. When someone is widowed, does their happiness tend to increase, remain level, or decrease?

[BlargMcLarg]

That's not enough. One of the often espoused counterarguments is "do happy people marry more, or does marriage make people happier?". You'd have to continuously check happy people and keep a control group from marrying. Checking just the widowers doesn't account for the severity of the loss.