The DIGNAD Model: Applications and a New Toolkit
Azar Sultanov, International Monetary Fund
Zamid Aligishiev, International Monetary Fund
Building resilience to natural disasters is imperative due to the rising threats posed by climate change, particularly for vulnerable developing economies. The DIGNAD (Debt-Investment-Growth and Natural Disasters) model was developed to analyze macro-fiscal implications of climate shocks and the role of economic policies in mitigating associated risks. DIGNAD is a small open economy model that allows users to assess debt sustainability risks linked to major natural disasters while explicitly considering the need to rebuild public infrastructure. Its rich general equilibrium structure allows for the construction of counterfactuals as well as broader scenario analysis involving ex-ante policies such as investing in resilient infrastructure, increasing fiscal buffers, and improving public investment efficiency. The recently developed DIGNAD toolkit offers a user-friendly interface, making the model accessible to users with limited experience coding with Dynare.
Published: 21 Oct 2024
A usual disclaimer applies that these are our views and not necessarily the views of the IMF, its management, or executive board. Unlike the earlier presentations by Junior and Costas, our project is much less technical, mostly revolving around making a relatively simple yet detailed representative agent model, easy to use by users without much experience in structural modeling. So we'll be presenting this toolkit today.
We are all aware of the dangers posed by climate change. An important one of which is an increase in the frequency and magnitude of natural disasters. And as demonstrated by recent events in Florida, which was hit by three major hurricanes within a span of two months, natural disasters can have a devastating effect, leading to loss of life, destruction of homes and livelihoods, as well as impacting critical public infrastructures. These effects can be persistent and require substantial government investment in reconstruction.
This can be particularly difficult in low-income countries and developing economies, which may have limited investment capacity and where increases in government borrowing could even threaten their sustainability. So building resilient infrastructure capable of withstanding such events is therefore essential, especially in such vulnerable countries. However, policymakers face a difficult challenge funding such investments in necessary climate-resilient infrastructure while keeping that at sustainable levels.
What we will present today is a toolkit that Azar Sultanov, Cian Ruane, who is now at the Central Bank of Ireland, and myself have developed while at the IMF's Research Department. And this toolkit is called DIGNAD. This is a toolkit that we believe can help precisely analyze these challenges and design policy scenarios.
The toolkit was available to Fund staff internally for quite some time and is published externally as of last year. We also provided an extensive user manual alongside the toolkit, which contains a lot more detail than what is included in today's presentation.
What I think is one of the key strengths of this toolkit is its user-friendly Excel-based interface, which we hope will boost its adoption by interested parties. Importantly, MATLAB proved to be a natural platform to combine a dynamic structural model with this user-friendly front end. It's an incredibly flexible and transparent platform for us to work in.
So the plan for today's presentation is that we will first present the highlights of the model that are relevant to the case of natural disasters and the macro fiscal implications. Then we'll discuss some of the applications that we've done so far. And finally, we will present the structure of the toolkit. And then, I guess, we'll have a couple of minutes for a Q&A session. I'll stop here and I'll leave the floor to Azar, who will discuss the model structure.
So thank you very much, Zamid, and thank you very much all MathWorks and all MathWork community to inviting us to this conference. So I will be quick actually to begin.
What is the DIGNAD? Well, DIGNAD stands for Debt, Investment, Growth, and Natural Disasters. It's an extension of the Debt, Investment, and Growth model, which has been developed in the research department and used at the IMF for many years. DIGNAD is tailored to disaster-prone countries. These are typically small countries or low-income countries that are particularly exposed to large climate shocks. In other words, countries with shocks that can disrupt the entire economy are frequent.
However, it can also be relevant for larger countries that will potentially be exposed to extreme climatic disasters in the future. The DIGNAD toolkit enables users to evaluate the macroeconomic impact of the variety of investment programs before and after a country is hit by a natural disaster, especially investment programs focused on climate-resilient infrastructure. The toolkit allows users to analyze the trade-offs involved in ex-ante climate adaptation investments compared to the need for ex-post reconstruction.
The DIGNAD model has been widely used in the Fund over the last five years as a part of Article IV staff reports, climate macroeconomic assessment programs, and most recently Article IV staff reports focused on application for the Resilience and Sustainability Trust. And as you can see, from the last year, the toolkit is also open to the wider public. And for those of you who are interested in learning more about the models in the short term, I recommend reading the two papers cited below, which covers in detail both DIG model of Buffie and others, and the DIGNAD model of Marto, Papageorgiou, and Klyuev.
So about the model structure, DIGNAD shares the same core structure as the DIG model, which can be summarized in three blocks. The policy block collects a set of fiscal and debt instruments that covers investment plans and debt financing. The private supply block contains firms' decision on labor and capital demand, which are affected by government spending and financing needs. And strategies-- the private demand block contains households' consumption and saving decision, which in turn are affected by government and firms' choices.
The dynamic, general equilibrium nature of the model ensures that these three blocks are interdependent and macroeconomic outcomes are jointly determined by the interaction. So let me start by going into a bit more detail into what is in the private demand block.
So one of the important features that the model is designed to capture is that many households in the emerging market and developing economies don't have access to financial instruments like savings accounts or credit cards. To capture this, the model features two types of households-- those that are called rule-of-thumb households or liquidity constraint households. These households can't save or borrow. And so they basically spend whatever income they receive each period.
The second type of households in the model are savers. And these are households who have access to various financial instruments and can invest in government bonds or in private companies. So that's who the different households are. But what they do?
Households work and earn labor income. They then use this income to consume domestically produced goods and services, goods produced abroad. The savers in the economy can choose how much of their income to invest and save for future consumption.
So what about the private supply block? Well, private firms can operate in either the tradable sector or the nontradable sector. Depending on the country, the tradable sector might be a goods producing sector, like agriculture, exports, or commodities. Or it could also be a tourism sector.
So what do private firms do in each sector? Well, they have a standard production function shown on the right. And they maximize their profits by choosing the optimal mix of private capital and labor. However, firms are more productive when there is a large stock of public infrastructure. You can think about this as firms needing to use roads, bridges, or other public infrastructure, for example.
An important feature of the model is that the total stock of public infrastructure is comprised-- composed of two types of-- subtypes of public infrastructure-- standard infrastructure and climate-resilient infrastructure. I will come back to the differences between these in a couple of slides.
Now, so lastly, let's consider the policy block. So the government in the model has different fiscal instruments. It controls the consumption taxes and labor taxes that households face, which is the main source of the revenue. However, increasing taxes, of course, reduce consumption.
Two of the most critical things that the government has to decide is how much to invest in standard infrastructure and how much to invest in climate-resilient infrastructure. The government can finance such investment using a variety of debt instruments, such as domestic debt, concessional debt, or external private debt. Lastly, the government may have access to some other external financing sources, such as grants from donors.
And when doing scenario analysis using the model, there are two ways of thinking about what the government does. We can assume it follows fiscal rules, which means that taxes adjust automatically to close the fiscal gap, or prevent excessive accumulation of debt. Alternatively, the user can assume that not such automatic adjustment takes place and that the government borrows to finance the fiscal gap, in which case they can see that would happen to the gross debt and fiscal revenues in the scenario.
So just to recap, the model features these three blocks. And importantly, the dynamic, general equilibrium nature of the model ensures that these three blocks are interdependent and macroeconomic outcomes are jointly determined by the interaction.
Now, let's move on the main novelty of the model-- natural disasters. These can affect the economy via several channels. The first, and one of the most important, are damages, which directly reduce output. These aggregate impacts can be decomposed into multiple channels.
The first channel is that disaster reduce the private capital stock. This is just due to the direct impact of natural disasters and means that in order to get back to its predisaster level of private capital, the private sector is going to need to invest in reconstruction. The second channel is direct disaster damages to public capital, such as roads or bridges being destroyed.
The third channel is the temporary productivity loss of private sector firms. This may be due to the requirement to clear debris or workers not being available for work as they rebuild their homes. An additional channel, which may be relevant in some circumstances, is a loss of creditworthiness. This makes it more expensive for countries to borrow from external private parties and therefore may increase pressure on existing other forms of relief or concessional debt to finance reconstruction.
The last channel is a decline in public investment efficiency during the reconstruction phase. You can think of this as being due to capacity constraints. Public investment efficiency is likely to be lower when a huge amount needs to be done at once. And management and monitoring resources may be extremely stretched.
So now into the second main novelty of the dignity model-- the two forms of public infrastructure we talked before. So resilient infrastructure may differ from standard infrastructure in four main ways. Firstly, it may be characterized by greater durability, meaning it might have a lower depreciation rate. This could be, for example, because it's designed to survive the usual wear and tear of the seasons.
Secondly, and most importantly, it mitigates the damages incurred when a natural disaster hits. Thirdly, climate-resilient capital may deliver a larger return than standard infrastructure. This could be because there is a very low capital initial stock of climate-resilient infrastructure relative to standard one. And therefore, they're quite high, the marginal returns.
However, resilient infrastructure may be more expensive than the standard infrastructure, both because it requires more specialized skills to build but also because it may require more specialized, imported components. While the first three features can make climate-resilient infrastructure very attractive, its high cost may increase the financial burden for governments looking to invest in climate adaptation, introducing a difficult trade-off.
And I know that was a very fast overview of the model. But I want to continue the next session. And I would like to ask Zamid to show us some example application that have been done in the past. And then we'll take over to show you the toolkit itself. Floor is yours, Zamid.
Thank you, Azar. So as previously mentioned, we have done a number of applications over the past years using the toolkit. These have been published in a variety of Fund publications, such as IMF working papers, special issue papers that accompany staff reports. The model was also the core of the macro fiscal implication section of the two climate macroeconomic assessment programs that the fund has done in the past. So most recently, the toolkit has been used widely in the work supporting the Resilience and Sustainability Trust Facility, the latest IMF financing instrument linked to the climate change risks.
So the main exercise that we can use the model for is to simulate the impact of a one-off natural disaster. The size of the disaster is something that we can choose based on country-specific factors, such as historical experiences of economic losses from past disaster events, or the desire to simulate a potentially large disaster that could occur in the future but was unprecedented in the past.
What we can then do with the model is to simulate the path of the main macroeconomic variables relevant for us, such as GDP, public debt, fiscal deficit, and investment and consumption expenditure. Importantly, one of the benefits of having a structural model is that we can do these simulations under different hypothetical scenarios-- for example, do counterfactual analysis. What is the impact of investments and resiliency if undertaken at the same time as structural reforms, for example, where the structural reforms would improve the efficiency of public investment processes?
We recently applied the DIGNAD model to Rwanda, aiming to assess the impact of ex-ante adaptation investments on growth and public debt. The model was calibrated to reflect key features of the Rwanda's economy, including the trend growth rate, debt-to-GDP ratio, and public investment efficiency. Importantly, the toolkit allows you to calibrate the simulation to a particular country case, using a variety of macroeconomic indicators.
So we simulated the hypothetical once in hundred years flood to assume that it happens in 2028, with a 4% immediate drop in GDP, focusing on infrastructure damage. Three scenarios that we've considered was the baseline scenario that assumed no ex-ante infrastructure investments prior to the disaster. And as mentioned, the flood was assumed to take place in 2028, which, in our horizon of simulations, was year five.
The second scenario assumed 1.5% of GDP per year invested in adaptation infrastructure, so before the natural disaster shock. And it was financed by concessional borrowing and external commercial financing. The third and last scenario combined enhanced investments in resilience and adaptation with reforms. So we assume that 3% of GDP per year increase in investment, paired with public finance and investment management reforms. The internal assumption was that we've increased-- the reforms lead to an increase in public investment efficiency of 20%.
So using such scenarios helps illustrate the trade-offs between all different policy measures-- so infrastructure investment, disaster preparedness, or long-term economic resilience.
The figures show GDP paths on the left and public debt to GDP path on the right for the three scenarios. In the baseline, which is shown in the blue line, GDP drops 4% when the disaster hits in 2028 and recovers slowly. However, it stays below the predisaster level for 12 years.
In the adaptation-only scenario, which is scenario two, depicted using the orange line, GDP rises before the disaster, due to this additional infrastructure investments, which are productive. The disaster causes less damage. There's also GDP falls from a higher level essentially. But still importantly, GDP still remains below predisaster level in the long term, even with this additional boost in structural resilience.
In the third scenario, adaptation plus reforms, depicted in-- with the red line, GDP grows even faster initially, due to an even larger investment scale-up, as well as more efficient investments. And it quickly recovers above the predisaster levels after the disaster hits.
So on the debt side, results present a trade-off. In the near term, more investments in adaptation and resilience increased debt to GDP. However, investments and reforms under scenario two and three produced lower debt-to-GDP trajectory over the long term, since reconstruction is less costly and is more efficient. Scenario three, in particular, highlights the benefits of investing in resilience to climate change and structural reforms for debt sustainability, as it shows that they can produce an outcome with both a brighter growth outlook and a smaller reduction in the fiscal space following a major natural disaster.
We also must consider that the initial investment in resilient infrastructure would likely be more expensive than the cost of standard structures, as mentioned before, by Azar-- 25% on average, according to some estimates. When you factor in all of these extra costs, countries could face serious fiscal challenges trying to invest in resilient infrastructure. And so donors might have to financially support the country. Remember, we are mostly talking about low-income, developing countries here.
There is then a potentially complicated trade-off for donors should they help finance the initial investment in resilient infrastructure with the prospect of a much smaller postdisaster disbursement or wait for disasters to occur and sustain the high cost of reconstruction. So obviously, this is not only an economic problem but also a political one.
And the great feature of the new DIGNAD toolkit is that it allows users to automate this calculation of donor net savings and produce tables like the one presented on the slide. What we show in the table here are the results for when the donor savings calculation was done for the case of Samoa Climate Macroeconomic Assessment Program.
They showed the net savings to donors under different assumptions about the magnitude of the natural disaster hitting Samoa. These net savings are computed as the difference between the fiscal spending for postdisaster reconstruction under the baseline assumption and the equivalent cost under the resilient infrastructure assumption. What these findings show for this particular case is that even for donors, it would be more cost effective to finance investments in resilience ex-ante than sustaining higher costs for reconstruction ex-post. This isn't necessarily going to be true for all countries and in all scenarios, of course. But the toolkit will allow users to explore these questions for their own country contexts.
Also, another important feature of the toolkit, which we prize, is the ability to align it with an externally supplied path of GDP under a natural disaster. For example, if you have simulated-- if you have estimated using empirical time series methods the impulse response function for a natural disaster, you can then take that and align model projections with that empirical re-estimated path to be more consistent with reality.
So now that we've covered just a couple of applications of this toolkit, I will stop here. And I will leave the floor to Azar, who's going to cover the more technical side of how the toolkit is structured and is operated. Thank you.
Thank you very much, Zamid. I would like to say that this is-- and as Zamid said, this is not all in the application. It can be applied for different countries, for different hypothetical situation or real situation. And I will just continue about new toolkit. And then probably, we'll have a time to take some questions.
So DIGNAD is a dynamic general equilibrium model, useful for economists looking to tailor their analysis of macro fiscal impacts of natural disaster and investment in resilience. Despite the advantages of the dynamic general equilibrium framework, its use may be constrained by the need of knowing MATLAB and dynamic programming. The DIGNAD toolkit aims to relax this constraint by providing an entirely user-friendly Excel-based interface.
It has enhanced user friendliness for calibrating the model, defining disaster, and financing scenarios and shocks. In addition, new models are added to enhance the user experience and automate processes that country teams or users will find useful. The core of the toolkit is a DIGNAD dynamic general equilibrium model that makes use of MATLAB and Dynare scripts and functions.
The toolkit combines an Excel front end and a back end that allows user to run it entirely from Excel. The only requirement is MATLAB to be installed on the user's computer. All inputs for calibration, policy and disaster scenarios, exogenous shocks must be provided in an Excel spreadsheet. The model can be calibrated using country-specific macroeconomic indicators, many of which are easily available to users or country teams or country economists.
These indicators are collected into four groups and color coded for better user experience. The user must specify the financing scenarios and can additionally calibrate the size and timing of natural disaster and reconstruction and the various mechanisms through which they affect macroeconomic aggregates. So finally, the toolkit provides user with a possibility of imposing assumption regarding the projected path of some indicators. The exogenous series worksheet accommodates eight of these series in three options.
So let's say if the country decided-- or its target on the budget that country will invest some amount to adaptation or climate-resilient infrastructure or to standard infrastructure or to use this money, for example, to save in special natural disaster fund, you can specify all these in the exogenous series worksheet and with three different options, as I may show you.
So the core of the toolkit is the simulation tool that produce the dynamic pass of key macro indicators. It's a collection of programs written on MATLAB that read user inputs from Excel file, simulate the model, and store the results in the output spreadsheets. Beside the core simulation tool, the toolkit also includes additional modules.
One of them, Alignment module, Zamid talked. We have Realism module, which checks the robustness to degree of resilience of adaptation infrastructure. Zamid also talked about Donor Savings module, which automates donor saving calculations. And Public Investment Efficiency module can simulate public investment efficiency reforms or changes.
The graphical tool embedded into the toolkit provides the option to visualize the simulation results on the screen for an on-the-spot inspection by users. It delivers a user-friendly way to generate MATLAB graphs using the Excel-based interface. Users, however, can construct their own graphs using a simulation result that are stored both in MAT file in MATLAB and also in XLS Excel output.
And the graphical tool can also be leveraged to produce scenario comparisons by presenting simultaneously projections implied by different scenarios. The tool allows to combine up to three scenarios and up to 16 variables in a single graph. And this concludes the brief overview on the toolkit. Now, floor is yours Zamid.
Thank you. We went a bit beyond our time limit. But I just wanted to quickly mention that we are currently working with MATLAB to bring DIGNAD to MATLAB Online, packaging it in an even more user friendly and faster front end. So we wanted to express our thanks to the team at MathWorks for their support and especially to for his excellent work on this. Thank you. I'll stop here.