Experiments and surveys have different advantages and disadvantages. Whether to use an experiment, survey or both depends on the objectives of the project.

Experiments

A well-designed experiment gives information about causal relationships and its conclusions may be that changing one variable that was controlled in the experiment (e.g. from variety A of wheat to variety B) will result in a specific change to the response variable (e.g. its yield or susceptibility to mosaic virus).

Experiments are important parts of many research projects

The problem with experiments is that they are usually conducted within a specific environment that is highly controlled such as a crop research station. There is always a question of how well the results will generalise to the more varied environments in the outside world.

Surveys

A survey is intended to describe the real world. An experiment may be able to assess the optimum amount of fertiliser and irrigation to use when growing variety A of millet, but a survey is needed to find how much this variety is used by farmers in a region. If farmers rarely use this variety, the results of the experiment will be of no practical importance.

Surveys can be used to describe relationships between variables (e.g. between the variety of millet grown and the yields obtained by farmers) but great care must be taken with interpreting such relationships since causal relationships cannot be inferred from surveys (e.g. the expensive millet varieties may only be grown by farmers on large well-irrigated farms). Measuring other variables (e.g. rainfall, temperature, soil type and farm management practice in each farm) and performing a complex statistical analysis may allow the researcher to be fairly confident that a relationship will be causal, but there is always some doubt.

Relationships that are observed from surveys often lead to research questions that can be rigorously assessed in subsequent experiments.

Surveys are also useful for assessing the practical impact of changes that are recommended on the basis of experiments.

Experiments and surveys

Large projects often involve a mixture of experiments and surveys and they may even be combined in the same data collection exercise.

For example, an experiment may be conducted on a sample of farms with 3 varieties of rice being used on different plots of each farm. Each farmer may also be given a questionnaire about education, age, farm size and management practice and attitudes towards different varieties — survey data.