Context Along with their core mission of counting the US population, the United States Census Bureau gathers a wide range of economic data. This dataset covers 16 of their economic reports and surveys: - Advance Monthly Sales for Retail and Food Services - Construction Spending - Housing Vacancies and Homeownership - Manufactured Housing Survey (1980-2013) - Manufactured Housing Survey (Current) - Manufacturers' Shipments, Inventories, and Orders - Manufacturing and Trade Inventories and Sales - Monthly Retail Trade and Food Services - Monthly Wholesale Trade: Sales and Inventories - New Home Sales - New Residential Construction - Quarterly Financial Report - Quarterly Services Survey - Quarterly Summary of State & Local Taxes - Quarterly Survey of Public Pensions - U.S. International Trade in Goods and Services Content - The data csv is arranged in a long format, with the time_series_code column tying it back to the metadata csv. If you're trying to figure out what data is available, you'll want to start with the metadata. - Just over a third of the time series store error codes, usually confidence intervals, rather than actual values. The metadata for these time series will have values in the columns `et_code`, `et_desc`, and `et_unit`. - All of the dates are stored as complete beginning of the period dates, but all of the time series are at either monthly, quarterly, or annual resolution. Exact days and months are provided for convenience when aligning time series and so that you don't have to unpack period codes like 'Q22009'. - There may be many time series bundled under a given data category or description. For example, the largest category (taxes) contains dozens of types of tax categories, and each of those contains a separate time series for each state in the country. - Two of the error code time series have non-numeric values. To convert the values column into reasonable units you'll need to drop all entries equal to the string `Less than .05 percent`. - The data have been substantially reformatted from how they are provided by the Census Bureau. You can find the script I used to [prepare the data here][1]. Acknowledgements This data was kindly made available by the United States Census. You can find [the original data here][2]. If you enjoyed this dataset you might also like one of the [other US Census datasets available on Kaggle][4]. Inspiration - The [National Bureau of Economic Research's macroeconomic history of the United States][3] covers many similar time series, but before the census data was reported. Can you integrate it with this census data? This should allow you to generate many time series stretching from the present back to the 19th century. [1]: https://gist.github.com/SohierDane/2c1b36f653724fbc7d8f26501ef4b88d [2]: https://www.census.gov/econ/currentdata/datasets/index [3]: https://www.kaggle.com/sohier/nber-macrohistory-database [4]: https://www.kaggle.com/census/datasets