An essential part of the scientific process is to verify results by trying to replicate them. FRED can be helpful in this regard, and we provide a simple example of this today. Gregor Smith published a 2008 study entitled Japan’s Phillips Curve Looks Like Japan, demonstrating that when you draw a scatter plot of the inflation rate minus the unemployment rate of that country for a sample period of January 1980 to August 2005, the resulting graph, commonly called the Phillips Curve, actually looks like a map of Japan. We replicate this result in the graph above, as one can clearly distinguish the large Japanese islands, Tokyo Bay, and various other features. So, yes, we have replicated the original result.

But the replication process is not only about reproducing results. You also want to check how robust they are to various modifications. The obvious modification here is to extend the sample to a more-recent date. We show this below, and it looks like the various islands have merged. With respect to Japan’s geography, this could only have happened if the seas lowered significantly or if new volcanoes emerged in the sea. As we do not believe this has happened in recent years, we must conclude that the original published result is not robust.

How this graph was created: Search for “Japan CPI” and select the monthly series. Then search for and add “Harmonized Unemployment Rate Japan”—again taking the monthly series. Modify the units of the first series to “Percent Change from Year Ago.” Modify the second series by applying the formula -a. Modify the date range of the graph to 1980-01-01 to 2005-08-01 and the graph type to “Scatter.” For the second graph, extend the date range to a more-recent date.

Suggested by Christian Zimmermann.