While it seems intuitive that programming develops logical thinking, it may be the case that people who program already possessed that skill and programming merely reinforces it.
Indeed. Additionally, we need to consider that the performance of a self-selecting group of students taking a course voluntarily may NOT necessarily reflect a general trend that would be applicable to ALL students when a course is required.
In other words, even IF programming does help develop logical thinking in students who are interested in it, it does not necessarily follow that these performance gains would happen with all students.
We need only look at the history of geometric proofs in high-school curricula to see the large-scale failure of another attempt to teach logical thinking indirectly in a high-school course. From the mid-1800s until the late 1900s, a full-year course in Euclidean geometry with emphasis on proofs was a standard part of most high-school curricula in the U.S. Yet a number of studies done in the past few decades have concluded that the logical skills actually developed in such courses were nearly non-existent outside a small group of students, most of whom probably already possessed significant logic and abstract thinking skills before taking the course. (Many studies concluded that the majority of students left such courses with little to no abilities to actually do mathematical proofs, and -- more disturbingly -- they also left the courses with profound misunderstandings about the nature of logic.)
I'm NOT saying that teaching logical thinking is hopeless, but it requires a combination of a good teacher and good student engagement, as well as a curriculum that is not "dumbed down" to accommodate "the lowest common denominator" of student. (This was a problem that many critics raised about geometrical proofs -- that they were taught in a way to make them "accessible" to everyone, but in the process they were dumbed down to a point that they no longer taught critical thinking very well. And the exercises were boring to those students who actually had an aptitude for such things.)
In sum, even if the measured skills show real improvements (not just selecting students already good at these things), it may be quite difficult to extend such improvements to uninterested students required to take a course, or to all students of all ability levels. We have loads of data on such attempts to teach abstract thought from math curricula reform over the past century or so... it rarely works as intended.