What You Actually Get When You Open This Book
The Handbook Of Computational Economics Volume 3 is a reference collection that covers simulation methods, agent-based modeling, numerical techniques for dynamic economic problems, and computational game theory. It is published by North Holland as part of a series. The volume brings together chapters written by people who actually build these tools, not people who just cite them. That makes a real difference when you are trying to implement something rather than just understand the high-level idea. I have used this book repeatedly over the years when working on heterogeneous agent macro models. One specific problem comes to mind. I was trying to solve a discrete choice model with aggregate shocks using the methodology described in one of the later chapters. The algorithm looked clean on paper but the numerical integration was blowing up every time I increased the grid resolution past a certain point. The text barely mentioned it. I spent about two days debugging before I realized the issue was related to how the solver handled near-singular matrices at the boundaries of the state space. The fix was basically to add a small penalty term to the objective function and switch to a different root-finding routine for the outer loop. The book does not walk you through that edge case, and honestly that is pretty typical of these volumes. They assume you will run into these things yourself. The chapters vary in depth. Some are essentially long tutorials with enough code-like pseudocode that you can implement them directly. Others are more survey-like and skip the gritty implementation details. If you are looking for a single chapter that teaches you how to build a complete agent-based economy from scratch, this is not the place. What you get is dense, specialized material that is useful once you already know your way around the basics.
A few things most people miss about using this volume effectively. The numerical methods chapters assume comfort with linear algebra and optimization. If your background is mostly in econometrics and you have never written a solver from scratch, you will struggle through the first few chapters. That is not a flaw in the book, it is just a prerequisite. Another thing: the book does not come with code. None of it. You are expected to translate the algorithms yourself. I have seen people treat the absence of ready-made implementations as a reason to skip the volume entirely, which is a waste. The algorithms are the valuable part, not someone else's MATLAB or Python code that might be ten versions out of date. One counter-intuitive insight worth mentioning. Several chapters cover methods that sound computationally expensive but can be made very efficient if you reorder your solution steps. For example, solving a rational expectations model by iterating on the policy function is often slower than iterating on the value function and then recovering the policy afterward, but the book sometimes presents them in the opposite order. Reading across chapters rather than following the table of contents linearly helps you spot these differences. The section on approximate aggregation is another place where the standard approach has serious limitations. It works fine when heterogeneity is low and idiosyncratic shocks dominate, but as soon as you introduce meaningful cross-sectional dependence or network effects, the approximation errors compound quickly. I ran into this when trying to model a simple supply chain network where firms shared suppliers. The method from the handbook gave reasonable results at first but the error grew nonlinearly as the network densified. I ended up switching to a direct simulation approach for that piece, even though it was slower, because the approximation breaks down in a way that is hard to detect before you run the model. Common pitfalls when working through this material:
People tend to treat the numerical examples in the chapters as universally applicable. They are not. The parameter ranges, grid sizes, and convergence tolerances are tuned to the specific models the authors used. If you copy them directly into your own work without adjusting, you will either get convergence failures or converge to the wrong solution. Another frequent mistake is ignoring the computational cost tradeoffs. The book presents methods alongside their accuracy claims but is less emphatic about runtime. A method that gives you an extra decimal of precision might take six times longer. In practice you often do not need that precision. The volume is available through Elsevier and various academic library platforms. If you are affiliated with a university you likely already have access through a subscription. Otherwise you can purchase individual chapters or the full volume from the publisher's website. Some older chapters may also appear in repositories like SSRN or ResearchGate, but those are unofficial copies and the quality can vary. I would recommend this book if you are doing computational work in economics and need to go beyond what introductory textbooks cover. It is not a good pick if you are still learning the fundamentals of dynamic programming or basic numerical methods. In that case, you would be better off starting with something like numerical economics texts by Stokey and Lucas or standard computational methods books before coming back to this volume. The content is solid, the chapter authors know what they are talking about, and it has held up reasonably well over time despite being a few years old. Just keep in mind that it is a reference, not a tutorial, and expect to spend more time on the implementation side than the reading side.
Get the Full Details
