br Discussion One of the indicators of wealth

Discussion One of the indicators of wealth that received a high importance is the number of household appliances. Possible explanations for the importance of this parameter are that it signifies a high and stable income over a long g protein coupled receptors of time; this can be compared to the high importance placed on measurements of incomes during 2002 and 2005. Another variable that received high rankings in both importance measurements is whether households owned a refrigerator, ranked as the third most important parameter according to conditional importance. Similar to the number of appliances, this could be interpreted as a wealth factor and might signify a wealthy household. It is also interesting from the perspective of possible connections between electrification and LPG uptake. The open fire is used not only as a means of cooking but also as a food preserver in certain areas of Vietnam and China. Hanging the food in the ceiling above the fire, thereby exposing the food to smoke, enhances preservation. A refrigerator together with an LPG stove may provide satisfying results both in terms of cooking and food preservation; after purchase of a refrigerator, one of the benefits of the open fire would be reduced. However, the results presented in this paper cannot distinguish between these explanations. Because of an unmeasured potential confounder, it would also be difficult to use regression techniques to measure any possible causation; i.e. it is possible that regardless of any co-benefits, households that prioritize kitchen improvements invest in both refrigerators and LPG stoves if they have the ability to do so. This also holds for the characteristic of owning a rice cooker, despite showing up as important also for which households that would later obtain LPG, i.e. whether this signifies wealth, an unknown household characteristic, or whether rice cookers have another connection with LPG. Although, in the light of which type of other variables that where shown to be important, the interpretation of these appliances as signifying wealth seems plausible and possibly a preference order in which household obtain certain goods, as suggested by Heltberg (2005). This is further underscored through the partial dependence on income for different values of past incomes and the number of appliances as shown in Fig. 3. If there are signs of past high household income at low income levels, the probability of a household using LPG was substantially increased. These households may have had the opportunity to purchase an LPG stove in the past and despite lower current incomes, they continued to use LPG. However, the past income and appliances are indicative also for making future fuel switching (see Fig. 5), indicating the possibility that households need a stable high income in order to make an investment in an LPG stove. Indicators of wealth, apart from current income, are seldom to any large extent included in studies of fuel switching (van der Kroon et al., 2013). When they are, they are sometimes interpreted as drivers of fuel switching instead of being interpreted solely as indicators of wealth. Although the results in this paper do not refute such assumptions, the high importance of past income levels in combination with these wealth factors may support the interpretation that it is a stable economic situation for a longer period of time or a previous high income that enabled households to make the investment. This finding can also be compared to previous noted strategy among poor households to behave in a risk averse manner (O\'Keefe and Munslow, Understanding). Village average land, distance to town and collection rate variables were also found to be usable predictors of LPG usage. A household with constraints on liquid funds may have to consider the cost and benefits of an LPG stove before purchase. For a household that have access to collectable wood and with a low opportunity cost for time, the difference in cost between cooking with wood and LPG is larger than for a household that find it harder to collect wood or have to purchase wood from a market. It is possible that village average land, distance to town and collection rate all capture this effect when predicting LPG usage. Furthermore, this effect seemed to be less accentuated for higher incomes, see Fig. 4, i.e. the partial dependence of village average land and distance to town were not as drastic for high income levels, indicating that at a certain level of wealth and income the effect of cost of fuel wood collection becomes less important. Note however, that these location variables may capture not only effects regarding fuel wood collection but also level of access to LPG, social practices as well as further unknowns.