Business Analyst Demographic Data: Confidence Levels

[RobertStevens]

Can someone from ESRI answer this please.

The BA demographic dataset has a large number of data fields.

But none of the data seems to have any confidence levels specified (variance, standard deviation, ..whatever).

Am I to assume that, for example, when I see that a certain

census tract has 1113 people in the age range 20-24

that the data is precise? Yeah, right!!

Another way of asking this question is:

How many people were surveyed in order that these numbers can be estimated?

How is the total population for a given geographic unit estimated?

The American Community Survey 5 year datasets also provide certain demographic

data down to the census tract level, but their data is always accompanied

by uncertainty estimate equivalent to 90% confidence level.

What methodology are Dunn & Bradstreet using?

(The data does come from that source does it not?)

Thanks

Rob Stevens

[Jason_RobinsonRobinson]

Robert,

First I would suggest reviewing Esri's methodology/modeling techniques (http://downloads.esri.com/esri_content_doc/dbl/us/J10268_Methodology_Statement_2015-2020_Esri_US_Dem...)

As far as margins of error, or MOEs, are published in the ACS because that program is a survey. Esri does not conduct surveys to produce updated estimates of the US population. Those types of accuracy measures are the domain of sample
surveys. Please reference the 2010 study (http://www.esri.com/~/media/files/pdfs/library/brochures/pdfs/vendor-accuracy-study.pdf) to learn more about Esri's accuracy with respect to competing private data vendors.

Regards,

Jason R.

[RobertStevens]

Hi Jason

Yes, but I am not interested in a theoretical study of ESRI's methodology. It would take me a huge amount of work to use any such comprehension into what I need: confidence levels on the numbers.  ESRI (or their vendor(s)) are best placed supply that.

If I wished to use ACS error estimates then I might as well just use ACS data directly and not bother with BA.

When one downloads ACS data, each table is accompanied by the 90% cf on the value. Surely, since BA data is likely to be used in statistical analysis, the cf's should be part of the data, should they not?

Rob