DataDebrief: US Unemployment Rate since 1948

This is a quick response to a discussion on the Visual Analytics LinkedIn group regarding a Wall Street Journal (WSJ) Graph of 'Historical US Unemployment'.

The WSJ chose a heatmap where ordinarily a simple line chart would suffice and discussion revolved around whether the heatmap was successful in providing information to the reader anymore than a line chart would.

Consequently, I put together a quick line plot in python + matplotlib using the following data sources:

United States Department of Labour - Bureau of Labor Statistics - Historical Unemployment rate
List of US Recessions post 1947 on wikipedia - An extract from the "post depression" section. This may vary from the WSJ determined periods. Wikipedia also contains a number of US (un)employment graphs in the entry for Unemployment.

The following graph is a little different to the WSJ one with respect to:

WSJ chose colour bands for each percentage point, I have chosen 0-4 as good, 4-8 ok, 8+ a bit high, loosely based on the colours of the WSJ example. I have no idea if these values are considered good or not by economists.
Periods of recession are represented by vertical bands, as opposed to white dots in the WSJ example.
For readability, the dates have been labeled every 4th year (as per Bureau) - it would vary depending on zoom level if it was implemented as a dynamic chart for the web.

Graph of US Unemployment rates 1948 - 2009

Graph of Australian and US Unemployment Rates 1948 - 2010
Note: Australian data only available from late 70s, source: Australian Bureau of Statistics - Labour force data

Code:

import datetime
import pylab
import matplotlib.pyplot as plt
import matplotlib.dates as mdates
from matplotlib.ticker import FuncFormatter as ff
from numpy import ma,logical_or, arange

rawdata = []
ymd = []

k = []
mdata = []

def chomp(string):
    return string.rstrip('\r\n')

f = open('SeriesReport20101014161301.csv', 'rb')
rawdata = f.readlines()
f.close()
for i in xrange(1,len(rawdata)): #skip header
    rawdata[i] = chomp(rawdata[i])
    yrates = rawdata[i].split(',')
    if i == 1:
        startyear = int(yrates[0])
    elif i == (len(rawdata)-1):
        endyear = int(yrates[0])+1

    for q in xrange(1,13):
        mdata.append(yrates[q])


for x in xrange(startyear,endyear):
    for g in xrange(1,13):
        k.append(datetime.date(x, g, 1))

fig = plt.figure()
ax = fig.add_subplot(111)
ax.plot_date(k, mdata, '-',label='unemployment rate', color='#571d1c', linewidth=1)
ax.set_xlim(datetime.date(1947, 12, 1),datetime.date(2010, 1, 1))
ax.xaxis.set_major_locator(mdates.YearLocator(4))
ax.xaxis.set_minor_locator(mdates.MonthLocator(1))

# acceptable? rate
p = plt.axhspan(0, 4, facecolor='green', alpha=0.05) 
#getting worse
p = plt.axhspan(4, 8, facecolor='#aaaaff', alpha=0.05) 
# bad?
p = plt.axhspan(8, 12, facecolor='red', alpha=0.05) 
#ax.xaxis.grid(color='white', linestyle='-', linewidth=0.5)

ax.set_ylim(0,12)
ax.yaxis.set_major_locator(pylab.MultipleLocator(1))
ax.yaxis.set_minor_locator(pylab.MultipleLocator(0.1))
#ax.yaxis.grid(color='white', linestyle='-', linewidth=0.5)

#recession http://en.wikipedia.org/wiki/List_of_recessions_in_the_United_States
plt.axvspan(datetime.date(1948, 11, 1), datetime.date(1949, 10, 1), facecolor='orange', alpha=0.1)
plt.axvspan(datetime.date(1953, 7, 1), datetime.date(1954, 5, 1), facecolor='orange', alpha=0.1)
plt.axvspan(datetime.date(1957, 8, 1), datetime.date(1958, 4, 1), facecolor='orange', alpha=0.1)
plt.axvspan(datetime.date(1960, 4, 1), datetime.date(1961, 2, 1), facecolor='orange', alpha=0.1)
plt.axvspan(datetime.date(1969, 12, 1), datetime.date(1970, 11, 1), facecolor='orange', alpha=0.1)
plt.axvspan(datetime.date(1973, 11, 1), datetime.date(1975, 3, 1), facecolor='orange', alpha=0.1)
plt.axvspan(datetime.date(1980, 1, 1), datetime.date(1980, 7, 1), facecolor='orange', alpha=0.1)
plt.axvspan(datetime.date(1981, 7, 1), datetime.date(1982, 11, 1), facecolor='orange', alpha=0.1)
plt.axvspan(datetime.date(1990, 7, 1), datetime.date(1991, 3, 1), facecolor='orange', alpha=0.1)
plt.axvspan(datetime.date(2000, 11, 1), datetime.date(2001, 10, 1), facecolor='orange', alpha=0.1)
plt.axvspan(datetime.date(2007, 11, 1), datetime.date(2009, 6, 1), facecolor='orange', alpha=0.1)


labels = ax.get_xticklabels()
for label in labels:
    label.set_rotation(30)
plt.legend(loc='best')
plt.xlabel('Date' )
plt.ylabel('% Rate' )
plt.title('U.S. Unemployment rate for the period 1948 - 2009')
plt.show()

Friday, October 15, 2010

US Unemployment Rate since 1948