To date, most studies have been cross-sectional in that they compare different areas with different food environments. Longitudinal studies are crucial because they provide valuable information as the research follows the same population over time. Natural experiments can provide good information, but it is important that the experiments are theory driven and nuanced by population and other variables. Longer time frames are often critical for judging the effect of different interventions and possibly linking a population’s food environment to its health. For natural experiments and other interventions, better surveillance methods can help researchers track information to see how an area is changing over time.
Policy and program interventions—such as those described in session five of the workshop (see Chapter 5)—were not generally set up by researchers. However, these activities may provide important opportunities for evaluation. To learn from both the successful and the unsuccessful elements, researchers could set up benchmarks for performance, sampling strategies, pre-testing of instruments, measurement of impacts on different sociodemographic groups, and process evaluations during interventions.
A theme that ran throughout the workshop was recognition of the complex physical and social environments in which food deserts are located. Approaches to understand some of the barriers can come from different disciplines working together.
The causal links between food deserts and health have not been firmly established. Researchers may need to look at more proximal behavior changes, such as shopping behavior, and then look at dietary behavior and ultimately disease outcomes and weight. Understanding the link between food availability and changes in obesity requires a better understanding of these intermediate steps, particularly the effect on dietary intake and shopping and eating behaviors.
Research shows that people do not adjust caloric intake when they consume calories via beverages. It is not understood why this is so, nor