How Do We Measure Age?
Despite the centrality of age, it is not as obvious as one may think to measure age. If someone states that they are 35 years old, this means they have already celebrated their 35th birthday, but not yet their 36th birthday. They are reporting what demographers refer to as their "age at last birthday". In that sense, age is always rounded down in demography, so that age reported as "35 years old" really means that the person is somewhere between 35.0 years and 36.0 years.
The measurement of age at last birthday can be contrasted with measures of "exact age". Exact age implies a more precise measurement, with details about the day of birth. It is often not explicitly stated whether a given age refers to age at last birthday or to exact age. For example, in analyses of children's mortality, talking about children aged 0-4 years old refers to their age at last birthday, whereas "under-5" refers to an exact age (all children who have not yet reached their 5th birthday). A lot of analyses focusing on the first year of life, in particular, will rely on exact age: when a child has not reached her/his 1st birthday, it makes little sense to study "age at last birthday".
Perhaps surprisingly, age is often misstated by respondents in censuses and surveys, or even during vital registration. In developing countries in particular, many people may not know their age because their date of birth was not registered, for example. But even in more developed societies, age misreporting during censuses or surveys can occur. It usually takes two forms: "heaping" and "shifting". By heaping, we mean that people tend to round their age to the nearest round number ending in 0 or 5. For example, if they are 37, they will declare that they are 40 or 35. Shifting is less clearly defined: in our societies where youth is valued, men and women may declare that they are younger than they really are; in other societies, the opposite may be true etc... Such errors are problematic because they distort the age pyramid of a population (see below). Demographers use various tools to detect such errors in censuses and surveys (e.g., Whipple's index).