We present evidence on evolving dietary patterns in Nigeria using six rounds of household consumption data from the Nigerian General Household Survey panel between 2011 and 2016. First, following conventional definitions in the literature, we show that Nigeria has not shown any aggregate increase in consumption of highly processed foods over this period, contrary to patterns observed elsewhere in the region. In fact, consumption of highly processed foods at home has declined, while food consumed away from home, often assumed to be highly processed, has risen substantially. We then show that estimates of food expenditure elasticities of different food types are highly sensitive to different estimation approaches and raise concerns about some frequently used methods in the literature. In the absence of credible exogenous variation, we argue for the importance of panel methods and household fixed effects to control for time invariant factors likely to confound cross‐sectional estimates. Finally, we examine semiparametric Engel curves for different food groups and find that apparent curvature in the relationships between food budget shares and overall food expenditure levels in the raw data become nearly linear when removing variation explained by time‐invariant household factors.

中文翻译：

---

尼日利亚加工食品需求的收入变化、不断变化的饮食和弹性估计

我们使用 2011 年至 2016 年尼日利亚综合家庭调查小组的六轮家庭消费数据提供了关于尼日利亚不断变化的饮食模式的证据。首先，按照文献中的传统定义，我们表明尼日利亚的消费总量没有出现任何增长。这一时期的高度加工食品，与该地区其他地方观察到的模式相反。事实上，在家中对高度加工食品的消费已经下降，而在外消费的通常被认为是高度加工的食品却大幅增加。然后我们表明，对不同食物类型的食物支出弹性的估计对不同的估计方法高度敏感，并引起了对文献中一些常用方法的担忧。在没有可信的外源变异的情况下，我们认为面板方法和家庭固定效应对于控制可能混淆横截面估计的时间不变因素的重要性。最后，我们检查了不同食物组的半参数恩格尔曲线，发现在去除由时间不变的家庭因素解释的变异时，原始数据中食物预算份额与总体食物支出水平之间关系的表观曲率变得接近线性。