Length of Numeric Fields

Hi there,

We currently have a survey in field and discovered that numeric questions sometime allow 10 digits to be entered, but sometimes only allow nine digits. For example you can input 2,000,000,000 but you cannot enter 3,000,000,000. This is a repeat survey from a few years back and in the last wave we were able to enter 10 digits always and now this time there seems to be this restriction. Is this an intended feature or a bug?

Thanks!

  1. your question must be checked as INTEGER. Uncheck INTEGER and it should accept more digits.

  2. re-think the questionnaire: In economics values of such magnitudes are not measured/reported with such a precision. The zeroes in your example confirm this. Yet the possibility of a mistake is great - and while collecting the tens and units, your interviewers will occasionally make mistakes in millions.

  3. If you care about the 2bln value - it stems from the 2^32 being the max value stored by 4 bytes long split by two for positives and negatives. In Stata terms:
    2bln
    This is slightly lower than the maximum in C#, since Stata reserves top 27 values for missing/extended missing values.

  4. For everyone else reading this: this is one of the reason you should NOT collect phone numbers as numbers, as they frequently overshoot this precision. Use the patterned text question of a kind: ### (###) ###-####

  5. I am also seeing countless surveys where novice designers multiply province codes by billions, district by millions , etc in an attempt to create a long unique id. Often trying to come up with 20-digit long IDs in countries with a some 5 thousands EAs. That is definitely not the proper way.

Regards, Sergiy Radyakin

Hi Sergy,

Thanks for the explanation. In future will take your advice of unticking integer and ask field teams not to use the 1dp.

My example with all the zeros was just an example trying to work out where 10 digits stopped being allowed. I agree that ordinarily we should not collect integers this long but in this project we are collecting financal data in Khmer Riels (taken from paper financial records so accuracy is important) and a few of the largest values are above this threshhold. Last round with this questionnaire (circa 2016) the data collected was of good quality. Recording the data this way was the preference of the field teams and is more accurate than having them divide totals by 000’s or converting to USD as they are copying directly from documents.

Good advice re: phone number and region codes.

Thanks!

Lachlan