In response to a recent paper by Chetty et al. showing that lowering benefits to tax sheltered retirement account does not have much influence on retirement savings Ray Fisman in Slate and other economists on twitter have asked if its time to end the 401k.
As Fisman states in his article "But getting rid of the tax shelters that burn a $100 billion hole in the government balance sheet might be one reform both sides can get behind"
I haven't read the Chetty paper fully, but I think a closer look might be necessary. What the paper does is estimate the influence of reducing, but not eliminating the subsidy for tax sheltered retirement accounts. The paper uses data from Denmark, which has similar tax laws but much better data availability. The authors find that for each dollar the government increases taxes on retirement there is a less than 1 cent reduction in total savings. In other words raising taxes on retirement savings doesn't change savings much. Driving this result is the fact that 85% of savers do not respond to changes in tax laws by adjusting their savings.
I was thinking about this in terms of the idea of a nudge vs a shove. If tax rates are adjusted just a little bit on savings, my guess is that most people wouldn't change their habits (consistent with the cited paper). If the changes are drastic those 85% passive savers might start to make adjustments. Furthermore, if tax breaks are eliminated for retirement savings then lazy savers might not start saving in the first place.
Many Americans use rules of thumbs to save 10-15% in their tax sheltered retirement accounts, if we eliminate the tax shelter totally the rule of thumb might change and you might see drastic changes.
The take away is behavioral economics is hard and a nudge of adjusting the benefit to retirement savings might produce a different result than a shove that eliminates the benefit totally.
That said I agree with Fisman and other economists that the tax break is mainly geared toward those with financial knowledge and/or wealth, so considering it might be worthwhile once we understand better the implications.