In the past, I created some graphs about our record low unemployment rate (US unemployment, and state-level unemployment), but does low unemployment also mean there are jobs available? Let's have a look at the data!...

Existing Graphs

I knew I couldn't be the only one interested in this kind of data, and therefore I turned to the Internet to see what might already be out there. And I found this really cool graph that Leonard Kiefer had posted on twitter (he's the Deputy Chief Economist at Freddie Mac, and he's always posting interesting graphs - you should follow him too!)

Getting The Data

I liked Kiefer's graph, so I decided to try to create my own version, and make a few small improvements. First, I located the data on the BLS website, and downloaded it. Where available, I selected the "seasonally adjusted" data. I now had 12 Excel spreadsheets (one for each graph). They all had their data in the same cells, therefore I wrote a re-usable parameter-driven macro to import this spreadsheet layout:

%macro read_data(xlsname,dataname,survey);
proc import datafile="&xlsname" dbms=xlsx out=&dataname;
range='BLS Data Series$a12:m32';
getnames=yes;
run;
data &dataname; set &dataname;
length survey $100;
survey="&survey";
run;
%mend;

I then used the following code to run the macro 12 times to import the 12 spreadsheets, and combine the 12 datasets into one:

%read_data(SeriesReport-20190411072035_0b99d7.xlsx, nonfarm, Total nonfarm);
%read_data(SeriesReport-20190411084943_987910.xlsx, ttu, %bquote(Trade, transportation, and utilities));
%read_data(SeriesReport-20190411091327_530482.xlsx, profbus, Professional and business services);
%read_data(SeriesReport-20190411091852_3ae78c.xlsx, eduhea, Education and health services);
%read_data(SeriesReport-20190411092515_a7430e.xlsx, leishosp, Leisure and hospitality);
%read_data(SeriesReport-20190411092906_c269ba.xlsx, manuf, Manufacturing);
%read_data(SeriesReport-20190411093420_c2726b.xlsx, gov, Government);
%read_data(SeriesReport-20190411094247_8b3456.xlsx, const, Construction);
%read_data(SeriesReport-20190411094638_8479b9.xlsx, serv, Other services);
%read_data(SeriesReport-20190411095010_9fd34a.xlsx, fin, Financial activities);
%read_data(SeriesReport-20190411095445_7d57c5.xlsx, info, Information);
%read_data(SeriesReport-20190411095643_4052a1.xlsx, minlog, Mining and logging);

data my_data; set nonfarm ttu profbus eduhea leishosp manuf gov const serv fin info minlog;
run;

My Preliminary Graph(s)

I did a bit of data-wrangling (transposing columns into rows, and such), and soon had the data structured so that I could create a grid of graphs using Proc SGPanel. Here's a summarized version of the code, showing the important bits, and a glimpse at my preliminary graph ...

proc sgpanel data=my_data;
panelby survey / columns=4;
series x=date y=job_openings_rate;
run;

My preliminary graph (above) just had a line, but I also want the area under the line to be shaded, and slightly transparent so you can see the grid lines behind it. Therefore I added a 'band' graph from the value 0, to the line.

band x=date lower=0 upper=job_openings_rate / fill transparency=.50;

The default colors are ~OK, but I liked the colors that Kiefer had used, therefore I set up an attribute table to map certain colors to certain graphs. I specified group= to turn colors on in the plots, and the attrid and dattrmap told the graphs which colors to use.

data myattrs;
length value $100;
id="some_id";
value="Total nonfarm"; fillcolor="cxf377e4"; linecolor=fillcolor; output;
value="Trade, transportation, and utilities"; fillcolor="cxff66b1"; linecolor=fillcolor; output;
value="Professional and business services"; fillcolor="cx629eff"; linecolor=fillcolor; output;
value="Education and health services"; fillcolor="cxde952a"; linecolor=fillcolor; output;
value="Leisure and hospitality"; fillcolor="cx11ae52"; linecolor=fillcolor; output;
value="Manufacturing"; fillcolor="cx2dbc9e"; linecolor=fillcolor; output;
value="Government"; fillcolor="cxc77dff"; linecolor=fillcolor; output;
value="Construction"; fillcolor="cxf77c73"; linecolor=fillcolor; output;
value="Other services"; fillcolor="cx20adee"; linecolor=fillcolor; output;
value="Financial activities"; fillcolor="cxbba21a"; linecolor=fillcolor; output;
value="Information"; fillcolor="cx83a414"; linecolor=fillcolor; output;
value="Mining and logging"; fillcolor="cx02bfc4"; linecolor=fillcolor; output;
run;

proc sgpanel data=my_data dattrmap=myattrs;
panelby survey / columns=4;
band x=date lower=0 upper=job_openings_rate / group=survey attrid=some_id transparency=.50;
series x=date y=job_openings_rate / group=survey attrid=some_id;
run;

My Final Graph

And the final step was to add the dashed reference line at the latest value! I added a new variable to the dataset, and only populated it for the last value in each graph, and then I plotted a dashed 'refline' at that value. But it wasn't visually intuitive that the dashed refline represented the last value in each graph, therefore I added a black 'dot' marker at the last value, to make it stand out more. (Here's a link to the full code, if you'd like to see all the details of creating the final graph.)

data my_data; set my_data;
by survey notsorted;
if last.survey then latest=job_openings_rate;
run;

proc sgpanel data=my_data dattrmap=myattrs;
panelby survey / columns=4;
band x=date lower=0 upper=job_openings_rate / group=survey attrid=some_id transparency=.50;
series x=date y=job_openings_rate / group=survey attrid=some_id;
refline latest / axis=y lineattrs=(color=gray77 pattern=shortdash);
scatter x=date y=latest / markerattrs=(color=gray77 symbol=CircleFilled size=8px);
run;

My Improvements

The original graph was nice, but I always try to make a few improvements. Here are the ways I think (hope) I've improved the graph ...

• The original graph plotted plain numbers along the left axis, and explained in the title that the values represented "(percent)". In my graph, I add '%' to the values (such as '2%') and therefore they need no explanation in the title.
• I make the background of the graphs a much lighter gray, which makes the colored area graphs show up more.
• I put a little more info in the footnote, indicating exactly when the data snapshot was taken.
• Also, when you click the footnote, it takes you directly to the data download page.
• I added a 'dot' marker to the latest value, so it's more evident that's what the dashed refline represents.
• And, I add HTML mouse-over text, so you can mouse over the lines, and see the exact monthly values. (Click the image of the graph above, to see the interactive version with mouse-over text). Below is a screen-capture example showing the mouse-over.

Now it's your turn - help me interpret these graphs! Do these numbers mean there are more jobs available now than in the past? Feel free to add your views/opinions/interpretations in the comments section.