Or, more accurately, why the left doesn't win all the time, as Meltzer and Richard would predict (see various previous posts if you're unfamiliar with this literature).
Just to recall the problem is, why don't the lower (L) and middle (M) income groups, who together form an unbeatable coalition to milk the rich (H), manage to coordinate? Or at least, often fail to do so.
Couple of insights on this. First, as Mark Blyth reminded me the other day, redistribution and the welfare state actually are not about milking the rich. On the whole, welfare capitalism is - and increasingly so - about lower and middle income groups agreeing to share income (through wage compression) and pool risk (through insurance such as unemployment coverage and pensions) amongst themselves. Sweden never had a particularly progressive tax system, and taxed capital gains lightly. So, the L+M vs M+ H dilemma presented by Iversen and Soskice misses the point.
Second, if Blyth is right and the question is all about L and M, then driving a wedge between these groups is far easier, since they have divergent interests if we exclude the possibility of milking the rich. So arguments about welfare spending benefiting L at the expense of M becoming all the more compelling.
The next step in completing the M+H coalition is the role of ideology in convincing M. Arguing that public spending is largely directed at the poor may be inaccurate, but most people aren't well enough informed to know how budgets break down. So propaganda can make a difference.
Here the debate about epistemic closure becomes interesting - even if M has a big interest in increasing government spending, the reluctance of some voters to willingly face the basic facts about redistribution subverts the process of social mobilization.
This probably makes very little sense except to me, but I'll post it anyway.