diff --git a/csvkit/cli.py b/csvkit/cli.py
index 415cd672b..2cf1a5ceb 100644
--- a/csvkit/cli.py
+++ b/csvkit/cli.py
@@ -412,7 +412,7 @@ def isatty(f):
         return False
 
 
-def default(obj):
+def default_str_decimal(obj):
     if isinstance(obj, (datetime.date, datetime.datetime)):
         return obj.isoformat()
     if isinstance(obj, decimal.Decimal):
@@ -420,6 +420,12 @@ def default(obj):
     raise TypeError(f'{repr(obj)} is not JSON serializable')
 
 
+def default_float_decimal(obj):
+    if isinstance(obj, decimal.Decimal):
+        return float(obj)
+    return default_str_decimal(obj)
+
+
 def make_default_headers(n):
     """
     Make a set of simple, default headers for files that are missing them.
diff --git a/csvkit/utilities/csvjson.py b/csvkit/utilities/csvjson.py
index a4356c4cb..afbe7af95 100644
--- a/csvkit/utilities/csvjson.py
+++ b/csvkit/utilities/csvjson.py
@@ -6,7 +6,7 @@
 
 import agate
 
-from csvkit.cli import CSVKitUtility, default, match_column_identifier
+from csvkit.cli import CSVKitUtility, default_str_decimal, match_column_identifier
 
 
 class CSVJSON(CSVKitUtility):
@@ -95,7 +95,7 @@ def main(self):
                 self.output_json()
 
     def dump_json(self, data, newline=False):
-        json.dump(data, self.output_file, default=default, ensure_ascii=False, **self.json_kwargs)
+        json.dump(data, self.output_file, default=default_str_decimal, ensure_ascii=False, **self.json_kwargs)
         if newline:
             self.output_file.write("\n")
 
diff --git a/csvkit/utilities/csvstat.py b/csvkit/utilities/csvstat.py
index c661c4d94..2a6342c0b 100644
--- a/csvkit/utilities/csvstat.py
+++ b/csvkit/utilities/csvstat.py
@@ -8,7 +8,7 @@
 
 import agate
 
-from csvkit.cli import CSVKitUtility, default, parse_column_identifiers
+from csvkit.cli import CSVKitUtility, default_float_decimal, parse_column_identifiers
 
 locale.setlocale(locale.LC_ALL, '')
 OPERATIONS = OrderedDict([
@@ -266,7 +266,7 @@ def calculate_stats(self, table, column_id, **kwargs):
                         op = op_data['aggregation']
                         v = table.aggregate(op(column_id))
 
-                        if self.is_finite_decimal(v):
+                        if self.is_finite_decimal(v) and not self.args.json_output:
                             v = format_decimal(v, self.args.decimal_format, self.args.no_grouping_separator)
 
                         stats[op_name] = v
@@ -352,7 +352,7 @@ def print_json(self, table, column_ids, stats):
         """
         data = list(self._rows(table, column_ids, stats))
 
-        json.dump(data, self.output_file, default=default, ensure_ascii=False, indent=self.args.indent)
+        json.dump(data, self.output_file, default=default_float_decimal, ensure_ascii=False, indent=self.args.indent)
 
     def _rows(self, table, column_ids, stats):
         for column_id in column_ids: