The results of a political poll are deep and complex. We tend to boil them down very simply though, because most of the time we're polling two-way, winner-take-all elections. Those races are generally zero-sum games between the candidates.
The Iowa Caucuses however are as far from a two-way winner-take-all election as you can get. They're a multi-way contest with proportional results. So the usual heuristics for reading polls are wrong. Contrary to rumor. Ted Cruz is not on the decline in Iowa, and has held steady for almost two months.
If you look at the leads in the indispensable Real Clear Politics chart, they vary a lot, from Cruz +10 to Trump +11 over the last month. This, combined with the volatility of the Cruz line in RCP's chart (which is caused by CBS/YouGov 40 reading which was grossly out of step with all the other polls), has made it appear that Ted Cruz had a bad January. But Cruz's numbers themselves, apart from that one outlier, have been remarkably stable.
To demonstrate it, I've plotted Cruz's Iowa figures for December and January, and a simple linear trendline for them in Apple Numbers. It turns out the slope of that line is -0.0000002, even if we include that 40 which is a ridiculous outlier. Effectively zero. Ted Cruz has stayed even:
Donald Trump's numbers are proving volatile. Eyeballing it, over the last few weeks he's won anti-Cruz supporters to his side, from the miscellaneous candidates at 5 and less. Ted Cruz though, he's got his support, and the polls don't really show it budging. Whatever he did to win over Ben Carson's supporters, he sure has held onto them.