Fix

Author: yuanx749

pandas/core/dtypes/dtypes.py

@@ -2344,6 +2344,8 @@ def construct_from_string(cls, string: str) -> ArrowDtype:
         if string == "string[pyarrow]":
             # Ensure Registry.find skips ArrowDtype to use StringDtype instead
             raise TypeError("string[pyarrow] should be constructed by StringDtype")
+        if pa_version_under10p1:
+            raise ImportError("pyarrow>=10.0.1 is required for ArrowDtype")
 
         base_type = string[:-9]  # get rid of "[pyarrow]"
         try: