Systematic error
Random error, as its name implies should not affect the central estimates (in statistical terms, the mean) of the variables subjected to analysis, although the measure of variation around that central estimate, and the precision of the estimate, are affected. As noted above, increasing the sample size is the easiest way to reduce the random error associated with sampling, while demographers tend not to pay much attention to random measurement error. Of much greater concern to demographers is the possible effect of systematic error in the collection of demographic data.
Systematic error, on the other hand, does affect the central estimate of the measurement as the errors (by definition) tend to run consistently in the same direction. As outlined in Section 3.2 
, systematic error can arise from both measurement and sampling. It thus follows that presence of systematic error from either source can distort seriously the results from a sample survey or census. These distortions are often referred to as ‘bias’. Importantly, bias can have distinct real-world implications for policy-making and intervention design. Ideally, one would like to remove as many sources of systematic error as possible before embarking on a data collection exercise. Equally, if systematic error is identified in data that have already been collected, this should be taken into account when drawing conclusions from the data.
The next sections describe systematic measurement error and systematic sampling error in turn.