[Piwoslaw]

I found this link at a local forum. It's a research paper comparing intersections with and without countdown lights.
The Effect of Digital Count-Down Display on Signalized Junction Performance
Here are the conclusions:

Following conclusions are drawn from the present research:
• The time to reach stop line for all car positions show the similar trend for count down the no-count down situations.
• There is a wide variation for the time to reach stop line, for the same car positions the reason may be attributed to difference in drivers’ abilities and car sizes.
• The discharge headway shows a high variance for the same car positions in different cycles for types of intersections. Although the headway is decreasing pattern for the subsequent cars in the queue but the trend very much differs from the theoretical trend presented in US Highway Capacity Manual, 1985.
• The digital timer has no significant effect on the travel time to stop line especially for the
first few vehicles.
• The digital timer has significant effect on discharge headway for all the cars in queue.
• The rate of red light violation is more for count down intersections than for non-count down ones.

(I highlighted the last point.)
The testing is far from ideal: The comparison was done for 3 intersections with countdowns and 3 without, so the difference may come from the intersections themselves and not from the countdown timers on the lights. A much better test would have been for the same intersections, with and without timers.

Has anyone found similar papers? A lot of European countries have displays on traffic lights and I haven't heard a bad thing about them, but I can't find hard facts. In fact, they were recently tested at 2 intersections in Wrocław and everyone liked them - both drivers and police. In Toruń countdown timers increased the number of cars going through the intersection by 70 cars per hour per lane. But I got this info from internet articles, not scientific research results.