Macroeconomic data are mostly continuous through time but exception exists. This is true for frequencies above daily data, but daily data there is almost never data for every day of the year.
Most utilizations of these data, being statistical or graphical except continuous data (or require extra work when the data aren’t continuous). It seems therefor preferable that the API serves continous time series at frequencies above daily data, while inserting a missing data code when necessary.
Making sure that the series are continuous can be done at the database level or at the API level.
An extra advantage of continuous time series is that it is then sufficient to know (communicate) the period of the initial observation and not the period of each observation.
- I would prefer that the time series be made continuous once for ever at the parsing stage
- I would suggest that, for frequencies above daily data, only the initial period of the time series be stored in the data base
- For daily data, it would be more efficient to store pair of period, value
- This requires to revisit every fetcher to make sure that the time series are continuous and adding necessary missing value codes in the data