John Friedman and Richard Holden have written this article for The New Republic Online (also available here), a popularized version of this working paper. The article argues: “We find that changes in redistricting have actually slightly reduced incumbent reelection rates over time. Instead, it looks like the other factors we mentioned above — money, polarization, media exposure, and perhaps others — account for the fall in electoral competitiveness in Congress.”
They further write:

The 1982 amendments to the Voting Rights Act–aimed at ensuring minority representation–have also curtailed the trickery of gerrymanderers. In particular, the preclearance provision has led to greater scrutiny of many districting plans. (This statute requires states with a history of racial gerrymandering–in practice, much of the South plus other areas with large Black, Hispanic, or Native American populations–to submit redistricting plans to the Department of Justice for approval prior to implementation.) Since race is one of the best predictors of party allegiance in many parts of the country, this has placed a further constraint on district manipulation. And, as one might expect, the incumbent reelection rate fell off in 1992, just after the first round of redistricting under this new law.

They conclude that the Voting Rights Act’s section 5 has played a crucial role in limiting gerrymandering. They conclude:

Now for what’s really scary: The sections of the Voting Rights Act that truly limit such practices are set to expire in 2007. And, though the independent commissions sought by activists might eliminate the ills of gerrymandering once and for all, support on the ground is still very thin– as voters in California, Florida, and Ohio have made clear. Furthermore, such work must progress state by state, so that complete protection by 2011 (when states must next redistrict) is highly unlikely.
Instead, these well-intentioned groups should direct their energies toward Congress and the reauthorization of the VRA. The VRA is fundamental in ensuring minority representation in Congress. Without it, gerrymanderers armed with sophisticated technology and facing few real constraints could run wild, causing minority representation in Congress to shrink dramatically. And that would be much more troublesome than a high incumbent reelection rate.

I will have to look at the academic paper in greater detail. But I’m wondering just from this description if it is really the non-expiring Section 2 of the VRA, not the expiring Section 5, that their model finds is responsible for curtailing the most egregious partisan gerrymanders.