1
$\begingroup$

I have a dataset of US firms (around 6500 in total) containing time series of:

  • Sales (191 negative observations corresponding to 0.0351% of total observations)
  • Liabilities (70 negative observations corresponding to 0.0129% of total observations)
  • TotalAssets (37 negative observations corresponding to 0.0068% of total observations)
  • Capital expenditures (5957 negative observations corresponding to 1.0947% of total observations)

What is the economic interpretation of those negative values? Are they just noise or reporting error? How to deal with them in an economic research?

$\endgroup$
2

2 Answers 2

1
$\begingroup$

Sales, liabilities, total assets and capital expenditures should be logically non-negative.

You should however consult metadata/manual for the dataset. Perhaps negative sales record situation where company had sales but customers asked for an return. There could be similar reasons why some dataset would record the other values as negative, but these reasons would break with typical definitions of these variables (such as the example above where returns would be recorded as negative sales illustrates). Without having access to metadata/manual of your dataset it is impossible to tell.

If there are no metadata/manual you can try to contact the person/institution who constructed the dataset for an explanation.

If it is impossible to contact people who constructed the dataset then you should probably get rid of those observations as they seem to be an error.

$\endgroup$
0
$\begingroup$

In many datasets missing values are recorded as -99 or -999.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.