Microsoft's certification exams serve one main purpose: to test a person's knowledge of Microsoft's technologies. As such, they are both too broad and too narrow.
They are overly broad in that you may use only a subset of these technologies in any one job. Each test has multiple areas of study, instead of having multiple certification tests, each on a specific subject. This has two side effects: any person taking the test must be somewhat familiar with more technologies than he or she may use in any one job, but at the same time it is not required to be thoroughly familiar with any of them. In fact, around 40% to 50% of the answers may be missed while still passing the exam. Is it important for an auto mechanic to have only a passing familiarity with both engines and body work, or is it better to specialize in the one or the other? Which of the three types would you hire to fix your engine? To repair a dent?
The tests are also too narrow for the purpose employers are using them: a "litmus test" of how "good" a programmer might be. A person can pass one of the Visual C++ exams without needing to understand C++ at all, much less be able to write readable, easily extensible, fast, stable, or bug-free (and bug-resistant) code. It doesn't test a person's ability to gather requirements or produce a good software design.
As a result of these shortcomings, many people think the tests are good for nothing. I would submit that the tests are good for one (and only one) thing: to test a person's knowledge of Microsoft's technologies (as stated above). In other words, they might be useful for a manager, not a programmer. They do not achieve the purpose many people think they do, to test how "good" a programmer might be. Unfortunately, while there are industry tests that cover the missing gaps, none of them seem to very popular. These include ICCP's ACP and CCP certification tracks, and IEEE's CSDP.