As discussed in a previous section, it is possible to run certain analyses of the micro data separately for each macro unit. Having linked the micro and macro data it is possible to do this in a way that is much more relevant to your research question. For instance, suppose you are interested in whether trust in politicians is higher where there is proportional representation (PR), you could estimate the mean level of trust in politicians where there is PR and compare it with the level where there isn't e.g. a dummy variable pr where 1= PR and 0 = not PR, so you could use the following Stata commands.

Example 6: Separate analyses for each macro unit

bysort pr: summarize trstplt

-------------------------------------------------------------------------
-> pr = 0

    Variable |       Obs        Mean    Std. Dev.       Min        Max
-------------+--------------------------------------------------------
     trstplt |       883    3.509626    2.264452          0         10

--------------------------------------------------------------------------
-> pr = 1

    Variable |       Obs        Mean    Std. Dev.       Min        Max
-------------+--------------------------------------------------------
     trstplt |      5282    3.883567    2.314891          0         10


or alternatively to get a sense of the variation between PR countries and then between non-PR countries, you could try the following command

Example 7: Variation between PR countries and then between non-PR


bysort pr country: summarize trstplt

---------------------------------------------------------------------------
-> pr = 0, country = FR

    Variable |       Obs        Mean    Std. Dev.       Min        Max
-------------+--------------------------------------------------------
     trstplt |       298    3.708054    2.148996          0          9

----------------------------------------------------------------------------
-> pr = 0, country = GB

    Variable |       Obs        Mean    Std. Dev.       Min        Max
-------------+--------------------------------------------------------
     trstplt |       298    3.348993    2.123862          0          9

-----------------------------------------------------------------------------
-> pr = 0, country = HU

    Variable |       Obs        Mean    Std. Dev.       Min        Max
-------------+--------------------------------------------------------
     trstplt |       287    3.470383    2.502794          0         10

------------------------------------------------------------------------------
-> pr = 1, country = AT

    Variable |       Obs        Mean    Std. Dev.       Min        Max
-------------+--------------------------------------------------------
     trstplt |       295    3.386441    2.351686          0         10

-------------------------------------------------------------------------------
-> pr = 1, country = BE

    Variable |       Obs        Mean    Std. Dev.       Min        Max
-------------+--------------------------------------------------------
     trstplt |       297    4.124579     2.15951          0          8


This style of analysis can be extended in various different ways. For instance, if the macro variable of interest is interval or ratio level (e.g. GDP per capita) the level of trust in politicians in each country could be graphed against GDP per capita with the country names as markers for the points. You could then subject the data for such an essentially macro level graph to a regression analysis; this is known as the two-step approach (e.g. Jusko and Shively 2005).

The University of Manchester; Mimas; ESRC; RDI

Countries and Citizens: Unit 4 Combining macro and micro data by Steve Fisher, University of Oxford is licensed under a Creative Commons Attribution-Non-Commercial-Share Alike 2.0 UK: England & Wales Licence.