Secondary sources are data produced by others. They can be contemporary or historical, qualitiative or quantitiative. They are usually already interpreted by someone else. They may provide data in appendices, but even that is presented data. We intend to find reliable data but it seems more and more people are involved in "spin" these days.
Written literature is itself secondary source material but it usually comes in the form of an argument. The better the notes and better the appendices, the more useful is such a book for you to check your argument against its. One has to be extremely critical of a presented argument, to see the links and the joins in the argument. You have to ask, "What could this person have said here?" Still it's good when someone has done the work and there is no real virtue in reinventing the wheel. If the work has been done, put it in a literature review and the bibliography, or use it in some way to support your argument, but try and go for the purer less argument subsumed data.
There are official statistics, and the biggest of them all is the census which dates from 1801. Official statistics are a huge and powerful resource base, especially for economic and social statistical surveys. The data is accessible and passes the practicality test. They are presented in ways that pass the reliability test - repeatable data and proper sampling frames. But is the data always valid? Does not the data contain lots of built in assumptions, lots of areas deliberately or innocently ignored, and in many cases optimistic presentation and spin? And then companies produce statistics, and so do all kinds of organisations like, for example, universities.
Of course a phenomenologist uses statistics in order to discuss the assumptions of government and the Civil Service. So perhaps researchers need something of a phenomenological view before a statistical re-analysis is made. We need to be aware just who produces statistics, official or otherwise. The law relating to Company Reports and auditing provide some check on wild claims but even there we have the best gloss presented on any situation.
There are government reports including those produces by public enquiries. Unfortunately while these reports contain valuable data the conclusions may in large part reflect the viewpoint of the Chair of the report and those carrying out the investigation.
Historical sources are important. However, as you go back in time such sources become more patchy in total and less rigorous in construction.
There are personal documents. For example, one day in the future someone could do research on me because I keep diaries. They could get a good insight into contemporary culture and issues as well as the life of a rather boring and unrepresentative individual. Which brings another point into focus with all documents.
This is to distinguish between witting testimony and unwitting testimony (Arthur Marwick). The witting testimony is where you reproduce an account of life that is directly given by that document and its authors. But there is also unwitting testimony where you read betweeen the lines. You get all sorts of built in assumptions showing through the given text, all sorts of background details. Sometimes a witting testimony can be in one direction, but the uniwtting testimony undermine what is formally written. It is up to the researcher to see the unwitting testimony as well as the witting testimony.
Then there is the study of the media and I'll include the internet in this. When you source the media you must include the time of transmission. So it is this programme seen then (unless you put it is on a tape). When you do a bibliography for the internet you must include the time it was viewed. Now the media is highly stylised and riddled with bias. Even where TV and radio programmes claim objectivity they ar packed with assumptions and the need for presentation. They are truncated and very selective. The internet is, fundamentally, one big advert. But even research sources there are concerned for the need for space. The meda may be practical as a source of data but it is likely to be highly research invalid and highly research unreliable - in other words it is baised and another programme is likely to say pecisely the opposite. And as for newspapeers, even broadsheet ones, they are very, very selective in what they choose to print and the limited data they present is usually heavily laden with opinion. Content analysis may well help you to tease out the bias in the media before it is used: this is where you analyse the presentation of the media for gender and class bias, for who is favoured and who is not, and you deconstruct the meanings presented. You must do this before making any generalised findings.
Best then to combine secondary and your own primary research if possible, or have ways of analysing the bias of secondary sources. This is often called triangulation, where secondary supports primary and primary supports secondary for accuracy, where qualitative research produces hypotheses for quantitative research and qualitative research shows where causality lies when one spots a statistical correlation within quantitative research because it is the case that quantitative research shows a snapshot of the big picture while qualitiative research shows a dynamic process at work. And nothing stops a mixing and matching of the two in any one piece of work.
Haralambos, M., Holborn, M., Sociology: Themes and Perspectives, London: Collins Educational, 748-754.