The properties of a Leontief production technology for Health System Modeling: the Thanzi la Onse model for Malawi
Martin Chalkley 1 , Sakshi Mohan 1 , Margherita Molaro 2 , Bingling She 2 and Wiktoria Tafesse 1 · 2025
Read The properties of a Leontief production technology for Health System Modeling: the Thanzi la Onse model for Malawi by Martin Chalkley 1 , Sakshi Mohan 1 , Margherita Molaro 2 , Bingling She 2 and Wiktoria Tafesse 1 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-21699.txt.
Introduction
Martin Chalkley and co-authors examine in 2025 the production technology inside a health system model for Malawi. The model treats healthcare production with fixed input proportions, a Leontief setup. At first glance, that looks restrictive beside general economic production functions, with constant returns and rigid corners.
Once placed inside an all-disease, agent-based model, those rigid traits vanish, they show. The framework becomes rich enough to describe real care production. That allows study of returns on health system investment in Malawi. The paper gives the model a firm conceptual base for future policy work.
Catalog record
| CREATOR | Martin Chalkley 1 , Sakshi Mohan 1 , Margherita Molaro 2 , Bingling She 2 and Wiktoria Tafesse 1 |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 14 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-21699
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
The properties of a Leontief production technology for Health System Modeling: the Thanzi la Onse model for Malawi Martin Chalkley 1 , Sakshi Mohan 1 , Margherita Molaro 2 , Bingling She 2 and Wiktoria Tafesse 1 Abstract As health system modeling (HSM) advances to include more complete descriptions of the production of healthcare, it is important to establish a robust conceptual characterisation of the production process. For the Thanzi La Onse model in Malawi we have incorporated an approach to production that is based on a form of Leontief technology – fixed input proportions. At first sight, this form of technology appears restrictive relative to the general conception of a production function employed in economics. In particular, the Leontief technology is associated with constant returns to scale, and level sets that are piecewise linear, both of which are highly restrictive properties. In this article we demonstrate that once incorporated into an all disease, agent-based model these properties are no longer present and the Leontief framework becomes a rich structure for describing healthcare production, and hence for examining the returns to health systems investments. JEL Code: I110, I120, I150, I180, L240, L300, L330. Keywords: Healthcare 1 Introduction Health system models (HSMs) provide the means to analyse and evaluate health system interventions ( Chang et al. 2017 ) . The Thanzi la Onse (TLO) model for Malawi ( Hallett et al. 2025 ) is a unique multi-disease whole system approach that provides a framework for simulating population health into the future accounting for the evolution of diseases and the impact of healthcare delivery. In developing this model we have been concerned with increasing the functionality of the elements that account for healthcare delivery and the link between available resources – healthcare workers, consumables and equipment – and the number of treatments that healthcare facilities can provide. This has led to developing a healthcare production framework in which we adopted fundamental principles from the economics of production ( Rasmussen 2012 ) . The concept of a production function is crucial to implementing this approach within the TLO model, and there are many choices of such functions. However, the data requirements for paremeterising most of these functions cannot by met from existing sources. We therefore focused on a comparatively simple functional form – the Leontief production function. In respect of the economic theory of production the Leontief function imposes very strong restrictions on the production process being modeled, including an assumption of constant returns to scale and the absense of substitution possibilies between inputs. The latter gives rise to (piecewise) linear level sets which are termed isoquants in production theory. Put simply the Leontief function is a (mostly) linear representation of the relationship between inputs and outputs. That would seem very unlikely to capture what we observe to be non-linear responses between resources and treatments in the model. However, as we explored the implementation of the Leontief function in the context of the highly multi-dimensional, multi-agent, dynamic and stochastic TLO model we discovered it is capable of capturing highly non-linear relationships between healthcare resources and treatments. This note sets out an explanation for how observable complexity can arise from a simple origin. 2 The simple one output two input Leontief production function We begin with the idea that the productive capacity of a healthcare system is determined by the resources it has at its disposal. In the simplest conception, we can consider a single measure of output – the number of treatments that are possible and denoted by y – as depending on two measures of input which for convenience we refer to as amount of healthcare worker time, denoted w and the volume of consumables, denoted c . The relationship between inputs and output is determined by what is termed the production technology which can be conveniently summarised by a mathematical function: y=f(w,c) (1) It is assumed that f , called the production function , is increasing in its arguments and quasi-concave, implying that its level sets, termed isoquants , are convex curves (viewed from the origin). There is particular interest in what happens to y as both inputs increase in equal proportions, this being the extent of returns to scale (rts). If output increases in proportion with all inputs there are constant rts and decreasing (increasing) rts if output increases in less (greater) than proportion to inputs. Figure 1 illustrates these properties graphically. In the left sub-figure the horizontal axis assumes all inputs increasing in proportion, and A shows how treatments increase if there constant rts, B shows the case of decreasing rts and C the case of increasing rts. In the right sub-figure the axes represent changing availability of the two inputs and the curves show combinations of those inputs that give rise to the same number of treatments. (a) Returns to scale: Constant (A), Decreasing (B) and Increasing (C) (b) Isoquants with substitutability Figure 1 : Graphical properties of a typical production function In order to offer as flexible an approach as possible, functional forms for f are specified to allow for variation in the shapes and positions of isoquants, and to allow for different returns to scale, both of which are determined through a discrete set of parameters. A number of popular functional forms have emerged as the basis for empirical examination of production relationships and one of the more general and frequently used of these is the constant elasticity of substitution function ( Solow 1956 ; Arrow et al. 1961 ; McFadden 1963 ) with a functional form of; f(w,c)=F(aw^{\rho}+(1-a)^{\rho})^{\frac{v}{\rho}}. (2) This functional form has four parameters; F (total factor productivity), a (factor share parameter), \rho (substitution parameter), and v (return to scale parameter). Although the CES function can accommodate subtle trade-offs between w and c , the requirements to provide reasonable estimates of the parameters are very demanding. This is true even when there is a single output and two inputs, but it becomes extreme when as in the context of an all-diseases HSM, there are many possible outputs and many inputs to consider. For the TLO model for Malawi, there is no way to estimate these parameters given current or any conceivably likely future data. A simpler conception of the relationship between input and output is to assume that each unit of output, i.e. each treatment, requires fixed amounts of healthcare worker time and consumables. This is the essence of a Leontief technology that we have adopted for the TLO model. In the two input, single output case this can be summarised in a a production function, y=\min\left(\frac{w}{a_{1}},\frac{c}{a_{2}}\right), (3) where a_{1} is the input requirement of healthcare workers for each treatment. To calibrate the input requirements a_{1} and a_{2} , there are a number of data sources that we can use, and these sources also relate to the different treatments represented in the model. The feasibility of the Leontief formulation for an HSM model comes at the expense of flexibility in specifying production. The differences are summarized in Table 1 below and depicted graphically in the two panes of Figure 2. CES Leontief Constant, decreasing or increasing returns to scale Constant returns to scale Non-linear isoquants Piecewise linear isoquants Substitution between inputs possible No substitution