Description
A need to increase the pandas version has been identified via the work in #885 that sought to remove the needs_pandas_0_22 and needs_pandas_0_17 decorators. For example, the function pvlib.irradiance.perez() specifically outputs different components whether it uses pandas>=0.18 or before and the tests are required to deal with this difference. From this perspective the goal is to simplify tests and avoid different functionality within supported versions of pandas.
A good guideline for the process of increasing versions is proposed by @wholmgren in #828 where he outlines the criteria as:
I suggest that we instead aggressively drop numpy/python versions when 1. there is a reason to do so and 2. the versions are outside the NEP-29 windows
The simplification of the testing suite and standardizing outputs certainly fall into 1. above as a criteria for deciding on the minimum pandas versions.