Clustering across Individuals/Households AND Areas

[Andreas Baltin]

Dear Statalisters,

I am currently working on a dataset in China regarding Individual spending. I have panel data and run something like the following:

xtset ID year
xtreg Spending HomicideRateCity PersonalIncomeCharacteristics Gender Age HouseholdCharacteristics, fe vce(XXXXXX)

Basically, I have a panel data set with numerous individuals. I have a measure for HomocideRate in the respective city the individual lived in at time t. I therefore want to cluster at city level. I think this should make sense. Then furthermore I have variables on individual characteristics, like height, age, gender, and personal income characteristics.
Then I have more characteristics across households: The survey basically asks things about the household, i.e. 'How much children live in this household', 'to how many cars does this household have access' etc. I also include these in my individual regression, where the values are all the same for each individual member of the household. From my understanding, I now also should cluster at the household level.

Is this thinking correct? --> I essentially want to cluster at City- AND at Household-level.

Many thanks in advance,
Andreas

[Andreas Baltin]

I have read this paper earlier: http://faculty.econ.ucdavis.edu/facu...009version.pdf

There they say: "For nested two-way or multi-way clustering one simply clusters at the highest level of aggregation. For example, with individual-level data and clustering on both household and state one should cluster on state. Pepper (2002) provides an example. If instead multi-way clustering is non-nested, the existing approach is to specify a multi-way error components model with iid errors.'

In my panel data set, no one moves city between the two survey round. Therefore my data should be nested right? --> Consequently I should just cluster on city, because this is a higher level than household? Does this make sense?