Dávid Krisztián Nagy教授分享经济地理学研究心得 Dávid Krisztián Nagy on Doing Economic Geography Research
本文最初于 2025 年 6 月 13 日 发布于微信公众号 Impactful Research；2026 年 4 月 28 日 同步至本网站。
Originally published on the WeChat official account Impactful Research on 2025-06-13; mirrored to this website on 2026-04-28.
来源：Google图文
这个公众号的第二十五篇文章，我们很荣幸邀请到西班牙国际经济研究中心的Dávid Krisztián Nagy教授分享他2018年发表在 Journal of Political Economy 的论文 The Geography of Development 的创作心得。
本文正文内容约九千字，全文阅读需约15分钟
#本期访谈主要问题
1. 写作的灵感与动机
2. 移民、技术传播和增长
3. 模型的反向预测能力
4. 未来的模型拓展
5. 写作中遇到的挑战
6. 文章未来的影响
7. 给予建议
Part 1: Inspiration and Motivation
写作的灵感与动机
Q1：What initially inspired you to undertake this research? I noticed that most urban economics modeling papers focus on either a single city or cross-city studies within a nation, whereas your work adopts a global perspective. Did theoretical literature primarily drive this, or did the idea emerge from observing real-world or historical patterns?
**
最初是什么促使您开展这项研究？我注意到，大多数城市经济学建模论文要么聚焦于单一城市，要么研究一国之内的跨城市问题，而您的研究却采用了全球视角。这是主要源于理论文献的启发，还是基于对现实世界或历史规律的观察？**
Yes, this paper originated from my collaboration with my PhD advisor–Esteban Rossi-Hansberg, and Klaus Desmet during my third year as a PhD student at Princeton University.
是的，这篇论文源于我在普林斯顿大学攻读博士第三年时，与导师Esteban Rossi-Hansberg以及Klaus Desmet的合作。
The core motivation stemmed from our broader interest in understanding how trade frictions and migration frictions shape economic growth, particularly the evolving geography of the world economy. Which regions will thrive in the future, and which will fall behind? From the outset, we aimed to develop a model that could explain the differential growth patterns across global regions.
研究的核心动机源于我们感兴趣的话题：理解贸易摩擦和移民摩擦如何影响经济增长，尤其是世界经济地理格局的演变 ——哪些区域将在未来崛起，哪些会落后？从一开始，我们的目标就是建立一个能够解释全球不同区域增长差异的模型。
Migration frictions naturally emerged as a key factor because they fundamentally constrain agglomeration, the spatial concentration of people and economic activity. Economic geography has long emphasized that agglomeration drives productivity and growth, so we hypothesized that migration barriers would significantly influence these dynamics.
移民摩擦天然地成为关键因素，因为它从根本上限制了集聚效应（即人口和经济活动的空间集中）。 经济地理学很早就指出，集聚通过提升生产率推动增长，因此我们推测移民壁垒会显著影响这一机制。
However, at the time (around 2012–2013), quantitative economic geography was still in its infancy. No existing model could directly address our question. The closest framework was the quantitative spatial model developed by Allen and Arkolakis (2014)[1], but it had two critical limitations for our purposes: (1) it assumed perfect labor mobility (no migration frictions), and (2) it was static, lacking growth dynamics. We realized we needed to introduce both endogenous productivity growth and migration frictions into the model. Fortunately, under our assumptions, the framework remained tractable and ultimately allowed us to explore these questions empirically.
然而，当时（2012–2013年左右）定量经济地理学尚处于发展初期。现有模型都无法直接回答我们的问题。最接近的是Allen和Arkolakis (2014)[1]提出的定量空间模型，但它存在两大局限：(1) 假设劳动力完全自由流动（无移民摩擦）；(2) 是静态模型，缺乏增长动态。于是我们决定在模型中同时引入内生生产率增长和移民摩擦。幸运的是，在我们的假设下，模型仍保持了可操作性，并最终为实证分析提供了框架。
[1] Allen, T., & Arkolakis, C. (2014). Trade and the Topography of the Spatial Economy. The Quarterly Journal of Economics, 129(3), 1085-1140.
Q2：**If we look at this question from an international perspective, cross-border movement can be highly restricted, as people are not always free to travel between countries. However, within a single country, such migration frictions tend to be much lower. So this reflects a real-world motivation behind the model or assumption, am I right?
如果我们从国际的角度来看，跨国流动往往受到严格限制，人们并不能自由地在国家之间迁移。然而，在一个国家内部，这种迁移的阻力通常要小得多。所以这就是模型或假设背后的一种现实动因，这样的理解对吗？**
Yes, absolutely. And what’s nice about our model is that it allows for any distribution of migration frictions across locations. So, for instance, you can incorporate varying levels of migration frictions, such as lower frictions within countries and higher ones across countries.
是的，完全正确。我们模型的一个优点在于，它允许在不同地点之间设定任意形式的迁移摩擦分布。比如说，你可以纳入不同层级的迁移摩擦——像是在国家内部摩擦较小，而在跨国迁移时摩擦更大。
Part 2:Migration, Technology Diffusion and Growth
移民、技术传播和增长
Q3：**Your paper suggests that, since we’re discussing a counterfactual scenario, under the current migration barriers, it would take around 400 years to reach balanced growth. So in the absence of large-scale migration liberalization as a substitute, do you think mechanisms like foreign investment or technology transfer could help accelerate this transition?
您的论文指出，根据这个反事实情境，在当前的迁移壁垒下，实现平衡增长大约需要 400 年。那么，在没有大规模移民自由化作为替代方案的情况下，您认为外国投资或技术转让等机制是否有助于加速这一转变？
Yeah, that’s a very good question. We incorporate that mechanism into our model as well, since we include technology diffusion. Admittedly, it’s modeled in a somewhat stylized way—we didn’t micro-found it through explicit channels such as foreign direct investment or the movement of ideas from one location to another. But we do include a general force of technology diffusion.
是的，这是一个非常好的问题。事实上，我们在模型中也考虑了这个机制，因为我们引入了技术扩散的要素。诚然，我们是以一种相对简化的方式建模的，没有从微观层面去刻画，比如阐述外国直接投资，或者思想是如何从一个地方传播到另一个地方的具体过程。但我们确实在模型中引入了技术扩散这股“力量”。
In the appendix of the paper, we conduct a robustness check where we increase the strength of technology diffusion to see its effects. And your intuition is absolutely correct: in a world where migration frictions remain at their current levels, increasing the strength of technology diffusion leads to faster global economic growth and quicker convergence of lagging regions toward the technological frontier. So in that sense, technology diffusion can act as a substitute for migration.
在论文的附录部分，我们做了一个稳健性检验，通过增强技术扩散的强度，来观察其对结果的影响。你的直觉完全正确：在当前迁移壁垒保持不变的情况下，增强技术扩散的确会带来全球经济的更快增长，也会加速落后地区向技术前沿靠拢的过程。 所以，从这个角度来看，技术扩散可以在一定程度上替代迁移。
However, we also find that this result doesn’t always hold. In particular, when migration frictions are relaxed, allowing people to move more freely, strong technology diffusion can actually reduce long-run growth. While it still boosts growth in the short run, over the long term, it can lead to slower growth. The reason is that when technology diffuses very easily across locations, individuals have less incentive to move toward high-productivity and high-density areas. In other words, if you can access the best technologies even in remote or low-productivity regions, there’s less motivation to cluster in economic hubs. As a result, we lose the benefits of agglomeration economies, which are crucial drivers of sustained long-term growth.
不过，我们也发现这种替代关系并不总是成立。尤其是在放宽迁移壁垒、允许人口自由流动的情况下，技术扩散在长期内反而会降低经济增长速度。虽然在短期内技术扩散仍然会促进增长，但从长期看，它可能会适得其反。原因在于，如果技术在各地之间扩散得非常容易，那么人们就会缺乏动力去迁往那些高生产率、高密度的经济中心。换句话说，即便你身处偏远地区，也能接触到最先进的技术，那就不再有强烈的动机去集聚在经济中心区域。这样一来，我们就失去了集聚经济所带来的增长红利，而这其实是推动长期持续增长的关键因素。
So there’s indeed a fundamental trade-off at play. While technology diffusion promotes convergence, it can undermine the agglomeration economies that arise from spatial concentration. In a world with freer population movement, we find that stronger technology diffusion actually reduces long-run growth. This counterintuitive result emerges because when knowledge spreads too easily across locations, it diminishes the incentive for workers to cluster together - thereby weakening the productivity benefits of density. So the relationship isn’t as straightforward as one might initially assume.
这里存在一个根本性的权衡。虽然技术扩散能促进区域收敛，但它可能削弱空间集聚带来的规模经济效应。我们发现，在人口流动更自由的环境中，更强的技术扩散反而会降低长期增长。这个反直觉的结果出现是因为：当知识可以太容易地在地区间传播时，工人聚集的动机就会减弱——从而削弱了密度带来的生产率优势。因此，两者的关系并不像最初想象的那么简单。
Q4 ：**What kind of policy or policy combinations can you imagine that can bring us faster to that good growth balance in the future.
什么样的政策干预（或政策组合）能够最有效地推动我们更快地实现未来最优增长均衡？
Our point in this paper is very clear: the most direct and impactful way to promote long-run growth and bring the global economy closer to its potential is by relaxing restrictions on labor mobility. Of course, we fully recognize that the political economy surrounding migration policy is complex. However, we firmly believe that enabling people to move freely—to the places where they wish to live and work—offers widespread benefits. This is largely due to agglomeration economies: when people come together in high-density areas, it generates positive externalities such as knowledge spillovers, shared infrastructure, and increased innovation.
我们在这篇论文中的核心观点非常明确：最直接、最有效的方式，可以推动全球长期增长、拉近各地区经济发展差距的，就是放宽劳动力流动的限制。当然，我们完全理解现实中与移民政策相关的政治经济问题非常复杂。但我们依然坚信，让人们能够自由地迁移——到他们希望生活和工作的地方——是一种能够惠及全社会的选择。这主要得益于所谓的集聚经济效应：当人们集中到高密度地区时，会带来知识溢出、基础设施共享、创新加速等正外部性。
This is not a controversial idea within economics. Scholars in economic geography and urban economics have been studying this mechanism for decades, and the empirical evidence is strong—greater population concentration consistently leads to higher output, greater productivity, and faster economic growth. So, from a policy perspective, if we are truly serious about addressing global disparities and unlocking growth, the most effective strategy would be one that targets the reduction of barriers to mobility. Unfortunately, we are still far from implementing such policies at a meaningful scale.
在经济学界，这一机制早已得到了广泛研究和证实。无论是经济地理学还是城市经济学，过去几十年的实证研究都表明：人口集中度越高，产出越高，生产率越强，经济增长也越快。因此，从政策角度来看，如果我们真的希望解决全球发展不平衡的问题，真正释放经济潜力，那么最有效的路径，就是减少人口流动的障碍。遗憾的是，目前我们离这一政策目标仍然相当遥远。
Q5 ：I think basically we’re moving forward in that direction because we see a lot of like visa-free policies between like China and some other countries and vice versa. So I think everything is going the right way.
我认为我们正朝着这个方向前进，中国和其他国家之间有很多类似的免签政策，反之亦然。一切都在朝着正确的方向发展。
Exactly. And hopefully, we’ll see more progress within countries, where the political economy is generally less complicated. Allowing people to move from one region to another within the same country is usually less controversial, both politically and socially. So I’m more optimistic that we’ll see continued progress on internal migration policies. Of course, across countries, the picture is far more complex. International migration involves more political sensitivities, institutional challenges, and coordination issues. That said, even small steps in that direction can generate significant benefits.
确实如此。我们希望在国家内部的迁移政策上能看到更多进展，因为相对而言，这方面的政治经济阻力要小得多。让人们从一个地区迁移到同一个国家的另一个地区，在政治和社会层面上通常争议较少。所以我对各国在内部迁移政策上取得持续进展持更为乐观的态度。当然，在跨国迁移方面，情况就要复杂得多了。国际迁移涉及更多的政治敏感性、制度障碍以及国家之间的协调问题。不过，即便是朝这个方向迈出小小的一步，也可能带来非常显著的经济效益。
Part 3:Backcasting Power of the Model
模型的反向预测能力
Q6 ：**I notice another very interesting result from the paper is that your model has successfully replicated the global population distributions in history, like from 1872 to 2000. Can you tell us how you achieved this, ang during this retrospective inspection, were there any historical details or regional patterns that particularly surprised you?
我注意到论文中另一个非常有趣的结果是，您的模型成功地复制了历史上的全球人口分布，比如从 1872 年到 2000 年。那么在这次回顾性检查中，是否有任何历史细节或区域模式让您特别感到惊讶。
Sure—let me explain this exercise a bit. What we do is calibrate our model using only current data, specifically, the current spatial distribution of population and economic activity across the globe. Then we perform what is sometimes called a backcasting exercise. Instead of forecasting future trends, we use the model to go back in time and estimate what the historical evolution must have looked like according to the model, to arrive at the present distribution we observe in the data. So, essentially, we ask: what dynamic process, as implied by our model, would have led to the current global economic geography?
我来简单解释一下我们在论文中所做的这个实验。我们对模型的校准仅使用当前的数据，也就是当今全球人口和经济活动在空间上的分布情况。然后我们进行了一项被称为“反向预测”（backcasting）的分析。不同于传统的前向预测（forecasting）——即预测未来会发生什么——反向预测的目标是：利用模型推演出过去可能发生过什么，从而产生我们今天所观察到的分布。换句话说，我们想问：根据模型的机制，是怎样的动态过程导致了今天全球经济地理格局的形成？
What really surprised me was how well the model performs—not just in replicating historical levels of population, but also in capturing regional patterns of population change over time. It accurately predicts which regions experienced faster growth and which grew more slowly, based solely on today’s data and the model’s structure. We run this backcasting exercise all the way back to 1870, and although the model fit naturally worsens the further back we go, it still performs remarkably well, even considering that this time span includes two world wars and other major historical shocks that are not explicitly included in the model. This result really convinced me of the model’s usefulness. The fact that it can match long-run historical changes so well, despite having no mechanical elements built in to guarantee this, suggests that it captures something fundamental about the evolution of the global economy.
让我感到非常惊讶的是，模型的表现非常出色。它不仅很好地拟合了历史上人口分布的绝对水平，更准确地捕捉到了不同地区人口增长速度的相对差异——即哪些地区增长更快，哪些增长更慢。我们将这项反向预测追溯到了1870年。当然，时间越久远，模型的拟合效果会有所下降，这是可以理解的。但即便如此，即使跨越了两次世界大战和其他许多模型未明确考虑的重大历史事件，模型的表现仍然令人惊讶地好。这个结果让我真正相信这个模型的价值。它并没有通过某种“硬编码”来刻意拟合历史变化，但仍然能很好地重现这些变化，这说明模型抓住了某种关于全球经济演化的基本规律。
After examining the fit across different regions globally, we find overall the model performs quite well everywhere. There isn’t any specific region that stands out as being fitted exceptionally better or worse compared to the global average. So, while the overall fit is strong, I don’t recall any particular case that notably deviates from that general pattern.
我们检查了模型在全球不同地区的拟合效果，总体来说表现都相当不错。没有哪个地区的拟合明显好于或差于全球的整体水平。所以，整体拟合效果很好，但我并不记得有哪个地区表现特别突出或特别逊色。
Part 4:Model Extension in the Future
未来的模型拓展
Q7 ：**Do you have any plans or suggestions for extending the model in the future? For example, are you considering incorporating factors such as climate change, artificial intelligence, or multinational supply chains into the framework? Could you also share some insights or hints on how researchers might build upon your previous work?
您未来是否有计划或建议来扩展这个模型？比如，您是否考虑将气候变化、人工智能或者跨国供应链等因素纳入模型框架？您能否分享一些思路或建议，帮助其他研究者基于您之前的工作进行进一步探索？
Yes, we have actually already incorporated climate change into our framework. We have two follow-up papers where we build on the same model to study two important questions related to climate change. In one of the papers, we use the original model to examine the impacts of coastal flooding expected in the future. This project involves collaboration with environmental scientists, not just economists. In another follow-up paper, we study the effects of rising global temperatures, which is one of the most significant impacts of climate change. For this, we extend the original model to include two sectors—agriculture and non-agriculture—since agricultural productivity is primarily affected by temperature increases. This work is done in collaboration with Bruno Conte.[1]
是的，实际上我们已经把气候变化纳入了模型框架。我们有两篇后续论文建立在同一个模型的基础上，它们研究了两个重要的与气候变化相关的问题。其中一篇我们使用了最初的模型来探讨未来可能发生在沿海地区的的洪水的影响。这个工作是和环境学家合作完成的，不仅仅是经济学家。我们的另外一篇后续论文研究了全球气温上升的影响，这是气候变化最重要的影响之一。为此，我们扩展了原模型，加入了两个部门——农业和非农业——因为农业生产力是受温度上升影响最大的领域。这项研究是和Bruno Conte合作完成的。[1]
Beyond climate change, we are currently working on another extension that incorporates the accumulation of human capital into the model. Human capital accumulation is an important source of economic growth that is currently absent from the baseline model. This is a challenging problem because it requires introducing additional dynamic processes—on top of the existing productivity dynamics—in the model. Specifically, we need to consider the trade-offs between the benefits of higher returns to skills and the costs associated with education and acquiring human capital. From a computational perspective, this extension is significantly more complex than the baseline model, which was surprisingly tractable despite modeling many heterogeneous locations. But we believe this extension is important because it allows us to study education policies and their effects. For example, if schools are built in developing countries, how does that affect human capital accumulation there? And with migration, do people stay where they acquire skills or move elsewhere afterward?
除了气候变化，我们目前还在研究另一项扩展：将人力资本积累纳入模型。人力资本积累是经济增长的重要来源，但在现有模型中尚未涵盖。这个问题非常有挑战性，因为它要求我们在已有的生产力动态基础上，引入另一套动态过程。具体来说，需要考虑技能回报的增加带来的收益和教育成本、获取人力资本的成本之间的权衡。 从计算角度来看，这一扩展比基础模型复杂得多，基础模型虽然涵盖了大量异质的地点，但在计算上还是相对可控的。而现在我们面临更复杂的动态问题。尽管如此，我们认为这项扩展很重要，因为它让我们能够研究教育政策及其效果。比如说，在发展中国家建设学校会如何影响人力资本积累？而且在存在人口迁移的情况下，这些人是留在积累人力资本的地方，还是学成后迁移到其他地方？
To our knowledge, there currently isn’t a global economic geography model with migration that can address these questions. So, despite the challenges, we believe this is a very interesting and valuable direction for future work.
据我们所知，目前还没有哪个世界经济地理模型能结合迁移问题来研究这些问题。因此，尽管困难重重，我们依然认为这是未来非常有价值且有趣的研究方向。
[1] Conte, B., Desmet, K., Nagy, D. K., & Rossi-Hansberg, E. (2021). Local sectoral specialization in a warming world. Journal of Economic Geography, 21(4), 493-530.
Part 5: Challenges
写作中遇到的挑战
**Q8: What was the biggest challenge during the research and writing process? Were there any unexpected obstacles during the submission and peer review stages for this paper?
在研究和写作过程中最大的挑战是什么？这篇论文在提交和同行评审阶段是否遇到了什么意想不到的障碍？
I think the biggest challenge was simply the scale of the model. While it turned out to be surprisingly computationally tractable, it’s still a large-scale model with many parameters. Calibrating and estimating these parameters took a significant amount of time and care. Gathering the appropriate data to take the model to the empirical level was also quite demanding.
Moreover, writing the paper itself posed challenges. When working with a framework of this size and complexity, it’s easy to get lost in the details—or even lose sight of the big picture. So, communicating the core ideas clearly in writing was something we had to be very mindful of. But fortunately, we were able to overcome those hurdles.
我认为最大的挑战就是模型的规模非常大。虽然最后我们发现它在计算上出奇地可行，但它依然是一个大规模的模型，包含大量参数。因此，校准和估计这些参数花费了我们很多时间和精力。此外，为了将模型真正应用到数据上，我们还需要收集大量合适的数据，这本身也是个不小的挑战。当你处理这样一个复杂且庞大的模型框架时，很容易陷入各种细节，甚至可能忽略掉整体的逻辑。所以在撰写论文时，如何清晰地传达核心思想，是我们非常在意的一点。幸运的是，我们最终还是克服了这些困难。
Q9 ：That sounds tough—especially in such a large project, it’s easy to get bogged down in the details.
确实很不容易，如此大的项目，有时候很容易在细节中迷失
Exactly, and not just in the details—sometimes even the broader narrative can get lost. That’s often the nature of large-scale quantitative work. But in the end, we’re happy we pushed through, even though it wasn’t easy.
完全正确，而且不只是细节，有时候连大的方向都可能模糊掉。这就是做大规模量化研究经常会面临的问题。但我们很高兴最终坚持了下来，虽然确实过程并不轻松。
Part 6: Future Impacts
未来的影响
**Q10 ：What kind of impact do you hope this paper will have in the future?
您希望这篇论文在未来产生什么样的影响？
So like I said, we already have a couple of follow-up papers. The two that I already told you about — one on coastal flooding[1] and the other on global temperature rise. We have a third follow-up paper as well in which we use, again, the original model to study the development of Asia in various counterfactual scenarios. This is a paper that came out in the Asian Development Review.[2]
就像我刚才提到的，我们已经基于这篇论文做了几项后续研究，包括一篇关于沿海洪水影响的文章[1]，另一篇则探讨了全球气温上升的经济后果。我们还有第三篇延伸研究，使用原始模型分析亚洲在不同反事实情境下的发展路径。这项成果已经发表在《Asian Development Review》期刊上。[2]
We are currently working on embedding human capital in the model. But we are hoping that there are lots of other extensions that people can do. And some of them they have already done. I mean, I know that some people have been using our framework to study other things. And one thing that seems quite timely is perhaps studying the effects of disruptions in trade. Our model is one that allows for any distribution of trade costs across locations, so you can change those trade costs and see what happens. And we live in a world now in which changes in trade costs are really on the table. I think that would be potentially another fruitful avenue.
我们目前正在尝试将人力资本的积累机制纳入模型。当然，我们也希望其他研究人员能基于我们的框架进行更多拓展，实际上已经有一些人在这样做了。我知道已经有学者用我们的模型研究其他课题。其中一个很及时的方向是探讨贸易中断的经济影响。 我们的模型本身就允许不同地区之间的贸易成本设定为任何数值，因此可以通过调整这些成本来模拟不同情境。现在这个时代，贸易成本的变化是非常现实的问题。我认为这是一个非常有前景的研究方向。
I think the key is that the model is really worth the time. It has all these heterogeneities — trade costs, migration frictions, differences in productivity, amenities, land — across locations. So there’s a lot that can be done with it. Luckily, it’s also computationally tractable. In fact, this is again based on Allen and Arkolakis(2014)[3]. We can characterize the uniqueness of the equilibrium under specific parameter conditions. We can offer a procedure that can be used to solve the model. And it always works. You don’t need to resort to complicated numerical methods. I think this is something that allows our model to be almost taken off the shelf and used by many researchers when they try to answer important questions.
我觉得关键在于这个模型本身非常值得投入时间去使用。它囊括了多个维度的异质性，比如贸易成本、迁移摩擦、生产率、便利性和土地资源等，所以它的应用范围非常广泛。幸运的是，这个模型在计算上也具有良好的可行性。事实上，这也归功于 Allen 和 Arkolakis(2014)[3] 的理论成果，使得我们可以在特定参数条件下刻画出均衡解的唯一性。我们提供了一套稳定的模型求解程序，始终有效，研究人员不需要依赖复杂的数值方法。因此，我们的模型几乎可以像“现成工具”一样被其他研究人员拿来使用，用于回答各种重要的政策和实证问题。
[1] Desmet, K., Kopp, R. E., Kulp, S. A., Nagy, D. K., Oppenheimer, M., Rossi-Hansberg, E., & Strauss, B. H. (2018). Evaluating the economic cost of coastal flooding (No. w24918). National Bureau of Economic Research.
[2] Desmet, K., Nagy, D. K., & Rossi-Hansberg, E. (2017). Asia’s geographic development. Asian Development Review, 34(2), 1-24.
[3] Allen, T., & Arkolakis, C. (2014). Trade and the Topography of the Spatial Economy. The Quarterly Journal of Economics, 129(3), 1085-1140.
Part 7: Advices
建议
**Q11 ：Could you please give us some advice to PhD students or researchers who are new to structural equilibrium models? How can they get started with their work, or how can they work efficiently with such models? Do you have any suggestions or steps that they might follow?
您能否给刚接触结构性平衡模型的博士生或研究人员一些建议？他们该如何开始工作，或者如何有效地运用这些模型？您有什么建议或步骤吗？
I think it’s always very important to start with the question rather than with the model. In the sense that maybe the question requires another type of model to answer, or requires only empirical analysis to answer. So the starting point should always be the research question. When empirical approaches prove insufficient, whether due to identification challenges or when evaluating hypothetical policies, we need to turn to structural modeling for counterfactual analysis. The choice of model should be guided by the research question; while not every case requires a quantitative spatial framework, when such models are appropriate, I strongly advocate starting with the most parsimonious specification possible. The process should indeed be gradual. Start with the simplest viable model, then carefully evaluate its limitations. If you find it inadequate, whether because it fails to match key empirical patterns or lacks essential mechanisms for your research question, that’s when you should consider extending it.
我认为研究问题本身出发而非模型出发，这一点始终非常重要。因为这个问题可能需要另一种模型来解答，或者只需要实证分析就能解答。所以，研究的起点应该始终是研究问题。当实证方法不那么有效时，无论是由于识别挑战还是在评估假设性政策时，我们需要转向结构模型进行反事实分析。模型的选择应以研究问题为指导；虽然并非所有案例都需要定量的空间框架，但如果此类模型适用，我强烈建议从尽可能简约的规范入手。这一过程确实应该是渐进式的。首先从最简单的可行模型开始，然后仔细评估它的局限性。如果你发现模型存在不足——无论是无法匹配关键的实证特征，还是缺乏研究问题所需的核心机制——这时才应考虑扩展它。
The key is to introduce new elements incrementally, one at a time. This disciplined approach allows you to precisely understand each additional component’s role in the model. The alternative—adding multiple features simultaneously—often leads to intractable complexity.
关键在于每次只逐步引入一个新要素。这种严谨的方法能让你准确理解每个新增组件在模型中的作用。如果一次性添加多个特征，往往会导致模型过于复杂。
These large-scale models already incorporate numerous mechanisms and dimensions of heterogeneity, with various forces pulling in different directions. If you introduce too many elements at once, it becomes extraordinarily difficult to isolate individual effects or even understand what’s truly driving the results.
这些大规模模型本身已包含众多机制和异质性维度，各种力量可能朝不同方向作用。如果同时引入过多新要素，我们将很难分离单个效应，甚至难以理解结果背后的真正驱动力。
In practice, I find it extremely valuable to study foundational quantitative spatial models through their seminal papers—works like Allen & Arkolakis (2014)[1], Redding (2016)[2], and others we’ve discussed. Many of these papers now come with replication packages available online, which presents an excellent learning opportunity.
在实践中，我发现通过经典文献来学习定量空间模型特别有价值——比如Allen & Arkolakis (2014)[1]、Redding (2016)[2]等我们讨论过的研究。这些论文大多配有在线的复制包，这提供了绝佳的学习机会。
I always recommend starting by attempting to replicate these published results. This process serves dual purposes: it helps you understand both the theoretical model structure and the numerical methods required to solve it. Through replication, you gain hands-on experience with the computational techniques while deepening your conceptual understanding.
我始终建议从尝试复制这些已发表的结果开始。这个过程有双重作用：既能帮助理解理论模型结构，又能掌握求解所需的数值方法。通过复制，你既能获得计算技术的实践经验，又能加深概念理解。
For those new to this field, I’ve written a review article titled “Quantitative economic geography meets history: Questions, answers and challenges” (published in Regional Science and Urban Economics) [3]. The paper develops a relatively basic quantitative spatial model framework and includes a simple numerical exercise using historical Hungarian data to demonstrate counterfactual analysis. All accompanying code is publicly available, making it particularly useful for educational purposes.
对于刚接触该领域的研究者，我曾在《Regional Science and Urban Economics》发表过题为”Quantitative economic geography meets history: Questions, answers and challenges”的综述文章[3]。文中构建了一个相对基础的定量空间模型框架，并使用匈牙利历史数据进行了简单的反事实分析数值演练。所有配套代码都已公开，特别适合教学用途。
This serves as another valuable entry point for researchers seeking to develop proficiency with quantitative spatial models. I consider this approach—starting with fully documented, replicable foundational models—to be the most natural and effective learning pathway.
这为想要掌握定量空间模型的研究者提供了另一个有效的学习起点。我认为这种从具有完整文档、可复制的基础模型入手的方法，是最自然且高效的学习路径。
[1] Allen, T., & Arkolakis, C. (2014). Trade and the Topography of the Spatial Economy. The Quarterly Journal of Economics, 129(3), 1085-1140.
[2] Redding, S. J. (2016). Goods trade, factor mobility and welfare. Journal of International Economics, 101, 148-167.
[3] Nagy, D. K. (2022). Quantitative economic geography meets history: Questions, answers and challenges. Regional Science and Urban Economics, 94, 103675.
学者简介：
Dr. Dávid Krisztián Nagy is a Senior Researcher at the Centre de Recerca en Economia Internacional (CREI), an Adjunct Professor at Universitat Pompeu Fabra, and an Affiliated. Professor at the Barcelona School of Economics. His main research interests lie in international trade, economic geography, and economic growth, with a focus on developing quantitative spatial models and integrating data to analyze the forces shaping the spatial distribution of economic activity. His work has been published in leading journals such as the Journal of Political Economy, Review of Economic Studies, AEJ: Macroeconomics, and AEJ: Microeconomics. His coauthored paper The Geography of Development received the Robert E. Lucas Jr. Prize. He currently serves as a Co-Editor of Regional Science and Urban Economics.
Dávid Krisztián Nagy博士是西班牙国际经济研究中心（Centre de Recerca en Economia Internacional，CREI）高级研究员，庞培法布拉大学（Universitat Pompeu Fabra）兼职教授，以及巴塞罗那经济学院（Barcelona School of Economics）客座教授。主要研究国际贸易、经济地理和经济增长，致力于开发量化空间模型并结合数据以分析经济活动空间分布的驱动因素。研究成果发表于 Journal of Political Economy, Review of Economic Studies, AEJ Macroeconomics, AEJ Microeconomics 等知名期刊。其合著论文The Geography of Development获得罗伯特-卢卡斯奖（Robert E. Lucas Jr. Prize）。目前担任Regional Science and Urban Economics 联合主编。
参考文献：
[1] Desmet, K., Nagy, D. K., & Rossi-Hansberg, E. (2018). The geography of development. Journal of Political Economy, 126(3), 903-983.
[2] Allen, T., & Arkolakis, C. (2014). Trade and the Topography of the Spatial Economy. The Quarterly Journal of Economics, 129(3), 1085-1140.
[3] Conte, B., Desmet, K., Nagy, D. K., & Rossi-Hansberg, E. (2021). Local sectoral specialization in a warming world. Journal of Economic Geography, 21(4), 493-530.
[4] Desmet, K., Nagy, D. K., & Rossi-Hansberg, E. (2018). The geography of development. Journal of Political Economy, 126(3), 903-983.
[5] Desmet, K., Nagy, D. K., & Rossi-Hansberg, E. (2017). Asia’s geographic development. Asian Development Review, 34(2), 1-24.
[6] Desmet, K., Kopp, R. E., Kulp, S. A., Nagy, D. K., Oppenheimer, M., Rossi-Hansberg, E., & Strauss, B. H. (2018). Evaluating the economic cost of coastal flooding (No. w24918). National Bureau of Economic Research.
[7] Nagy, D. K. (2022). Quantitative economic geography meets history: Questions, answers and challenges. Regional Science and Urban Economics, 94, 103675.
[8] Redding, S. J. (2016). Goods trade, factor mobility and welfare. Journal of International Economics, 101, 148-167.
| 责任编辑 | 张帆 |
| 整理翻译 | 张诗怡 |
| 校对 | Dávid Krisztián Nagy 张诗怡 |