Natural resource collection and desired family size: a longitudinal test of environment-population theories

Note, this prediction assumes that when natural resources diminish households cannot readily substitute other goods, such as purchased fuel, for those previously gathered. That is, the specific local social organization may condition this relationship. In poor, agrarian societies like Nepal this assumption is reasonable. However, in other settings, where markets are more accessible in terms of household finances and location, households may substitute alternatives for local forest products. This may allow households to support the same size or smaller family with fewer resources and children may be less desired because there is less need for their labor (see Brauner-Otto 2014 for more on this and other pathways and conditions).

We estimated all models with respondents who were under age 45 at baseline, and all effects were stronger than those shown here.

We also estimated models using an alternative measure for time of resource extraction that calculated the average time to collect a resource for all households in a neighborhood that gathered that resource. This measure then provides an estimate of the natural resource availability of the neighborhood, accounting for typical needs in the neighborhood. We then used that neighborhood average as the value of time to collect fodder or firewood if that specific respondent lived in a household that did not gather that resource. Results for fodder are substantively identical to those presented here, but the alternative measure of time to collect firewood was not statistically significant once we included any additional variables in the models.

A few households reported extremely large numbers of some of these animals (e.g., 23 female buffaloes and 24 sheep or goats). We top coded this variable at 15 as 98% of respondents reported 0–15 livestock total.

Of course, individuals and households elect where to gather their resources, and they likely are incorporating the quality of natural resources and their fertility desires simultaneously as they make that choice. Additionally, previous research in this setting has found that flora quality is itself related to fertility behaviors (Brauner-Otto 2014). Therefore, we also explore measures of the quality of the natural resources. Data come from flora plot surveys conducted in 1996 and 2006 from 265 flora plots selected from a variety of types of areas, located on the perimeter of the study site, and arranged at 250-m intervals along equally spaced (1 km) transects that extend 1250 m away from the perimeter (see Dangol and Maharjan 2012 for a more detailed discussion of the data collection). Flora teams counted the number of different tree, shrub, and grass species in each plot yielding measures of species density (the number of plants in a flora plot), richness (total number of species present in a flora plot), and the balance between the two (plant diversity with the Shannon-Weiner Diversity Index, an index commonly used in ecological research and is considered a measure of biodiversity (c.f. Chiarucci et al. 1999; Patil and Taillie 1982)). We examined measures of the closest flora plot (as measured by distance as the-crow-flies) to reveal the most immediate environmental conditions facing individuals and households, but following previous research on contextual effects in this setting, we also investigate geoweighted measures that incorporate the entire study area weighting those plots closest to the respondent more heavily than those farther away (see Brauner-Otto 2014). However, none of the measures of flora diversity, density, and richness were significantly related to a change in desired family size. This may be because of the specific time span these measures refer to.

Neighborhood measures come from Neighborhood History Calendars which cover up to 2003 and are lagged. Therefore, time 1 measures describe the neighborhood in 1995 and time 2 describes 2003. We also looked at measures of the number of years a respondent had a school, health service provider, market, or employer with a 5-min walk. These were not statistically significant in any model.

We estimated models excluding those who never gathered the resource and found results virtually identical to those we present here.

We also explored the role of migration but did not find any significant effects for whether a woman had moved from her baseline neighborhood. Other migration-related dynamics such as the labor migration of other household members may be a part of the overall household functioning, and future research should examine the interaction between household migration patterns, resource use, and fertility preferences and behaviors.

To create these measures, we first calculated the change in time to collect fodder and firewood. We then broke these distributions up into five roughly equal groups (quintiles) and created separate dummy variables for each category. Respondents were then coded 1 for the range that included the change in resource collection time they experienced. For example, if a women lived in a household that reported that they spent 1 h collecting firewood at baseline and 45 min collecting firewood at follow-up, she would be coded as 1 for “Decrease by 2 h to 5 min (firewood)” and 0 for all the other dummy variables.