python2 string formatting - old and new - different for unicode

[buran]

replying to another thread, I noticed something strange. following script, running python 2.7

import forecastio

def main():
    api_key = "API KEY"
    lat = -31.967819
    lng = 115.87718
    forecast = forecastio.load_forecast(api_key, lat, lng) 
    by_day = forecast.daily()

    print "===========Daily Data========="
    print "Daily Summary: %s" %(by_day.summary)
    
    print "===========Daily Data========="
    print "Daily Summary: {}".format(by_day.summary)

if __name__ == "__main__":
    main()

and the result

Output:
===========Daily Data=========
Daily Summary: Light rain on Friday through Monday, with temperatures falling to 15°C on Saturday.
===========Daily Data=========
Traceback (most recent call last):
 File "dark.py", line 53, in <module>
   main()
 File "dark.py", line 46, in main
   print "Daily Summary: {}".format(by_day.summary)
UnicodeEncodeError: 'ascii' codec can't encode character u'\xb0' in position 68: ordinal not in range(128)

so old-style string formatting has no problem with unicode char, while new one - using format, raise error. The behavior still the same if explicitly specify {:s}. I find it odd that there is difference. Do I miss something?