文 | 霞光智库，作者 | 陈玉宇，北京大学光华管理学院教授、北京大学经济政策研究所所长
发达国家早就是发展中国家的人工智能。汽车怎样造，电厂怎样建，世界上早有人知道，可是一百年来真正追上来的国家是少数。今天AI使知识更便宜，于是旧问题又回来了：便宜的知识过去没有自动变成增长，更便宜的知识为什么会？
这篇文章在增长理论里切开一个长期被压缩的环节：一群人怎样变成一个能共同使用知识的生产组织。作者提出”组织技术”和”双侧校准”两个概念，说明为什么生产能力可以在知识一页不少的情况下消失，为什么第一家工厂最难、第二家容易，为什么一个地区学会了生产未必学会了组织生产。这个角度被用来重看苏南、珠三角、温州、东北的工业化，福建一个村庄怎样成片学会房地产，以及两件正在发生的事：中国企业大规模出海，和制造业收缩以后新组织从哪里来。
文章最后回到AI。作者认为，AI对增长的作用取决于它改变的是知识供给，还是组织校准，还是两者都改变；同样的技术潜力，在校准快的行业和校准慢的行业会走出两条完全不同的生产率路径。
这是一篇思想长文，不是短评。它提出问题，综合已有认识，给出一个新的研究对象，并明确划定自己的解释边界。适合愿意花四十分钟认真读完的读者。
为什么世界已经拥有大量先进知识，许多国家却始终难以实现快速增长？发达国家早就是发展中国家的“人工智能”，今天AI又使知识的创造与获取更加容易。但知道怎样生产，与真正形成生产能力，是两个不同的过程。
本文综合增长经济学、技术扩散与组织经济学的已有认识，提出从组织技术的角度重新审视知识向生产率的转化。其中，“双侧校准”描述不同角色通过共同实践形成相互适配的工作规则。个人可以带走角色经验，已经形成的共同校准却不能由任何一方独自携带。这使生产组织具有独特的形成、积累、折损与重建规律。
这一视角有助于重新理解产业集聚、中国区域工业化、企业出海以及产业结构转型，也为解释AI增长效应的行业差异提供了新的思路。AI既可能增加知识供给，也可能缩短甚至改变组织校准过程。但现实增长还取决于资源重新配置和消费需求的发现。
生产知识的可复制性与生产组织的可复制性，是两种不同的技术性质。理解二者如何连接，是研究人工智能时代经济增长的重要方向。
知道一件事怎么做，和一群人把它做成，中间隔着一段路。
一、发达国家早就是发展中国家的人工智能
今天，人们对人工智能推动经济增长抱有很高期待。机器能够编程、设计产品、解释技术难题，也开始参与科学发现。知识生产的成本大幅下降，经济增长率会不会随之大幅提高？
这个问题并不新。回答它之前，不妨先看一场已经进行了一个多世纪的实验。对发展中国家来说，发达国家早就是它们的人工智能。汽车怎样制造，电厂怎样建设，药品怎样生产，现代工厂怎样管理，世界上早有人知道。设备可以买，工程师可以聘，学生可以送出去学习，跨国企业甚至可以把整套生产线搬过来。知识并非免费。知识产权、贸易限制和政治环境都可能形成障碍。但获得现成知识的成本，远低于重新发现这些知识的成本。
设想一个人在1960年作这样的推理：世界上的生产知识对加纳和印度几乎是现成的，先进国家已经付出了发现的成本，后来者只需要拿来用，所以这些国家会很快追上来。这一推理在当时并不荒唐。格申克龙讲“落后的优势”，讲的就是这件事。许多发展经济学家大体也持类似看法。六十多年过去，结果我们都知道。日本、韩国、中国等经济体的追赶引人注目，恰恰因为它们是少数。大量经济体长期接触先进知识，却没有形成与之相称的生产能力。发展核算的文献反复确认了这一点：各国人均收入的差距，大部分不能用物质资本和受教育年限解释，剩余的那一大部分被归入“全要素生产率”名下。
再做一个更极端的思想实验。假如今天把世界上最先进的生产知识完整地交给一个生产率很低的国家，技术图纸、生产流程、管理手册、全部科学知识，一样不少。它能不能在十年之内变成发达经济体？没有人会这样打赌。那么，它缺的究竟是什么？资本、制度、市场、基础设施，这些当然都缺。但还有一类东西，不在任何知识清单上，也不能靠交付获得。它只能在实际生产活动中慢慢形成。它是把知识、人员和资源结合成现实生产活动的能力。
这就是本文想追问的问题。我们对知识怎样创造、怎样传播，已经有很深的研究。我们对一个社会怎样获得使用知识的能力，认识还远远不够。知识的可获得性与知识的经济实现是两件事。把两者当成一件事，是那个1960年推理的全部错误。今天关于AI 的推理，很多正在重复这个错误。仅凭机器能够完成多少知识任务，就推算它能使GDP 增长多少，跳过了中间最难的一步。
AI 使这个旧问题变得更尖锐。过去知识已经便宜，现在知识要变得更便宜。如果廉价知识过去未能自动转化为增长，更廉价的知识为何就能？
还有一层。增长不是把知道的事做成一次就完了。收入提高，需求变化，旧的产业收缩，新的产业要长出来。一群人学会做一件事以后，社会还要不断地让另一群人学会做另一件事。这篇文章先讲前一半，第一家工厂怎样建起来。后面再讲后一半，旧工厂关了以后，新组织从哪里来。
二、知识还在，生产能力却可能消失
还可以从反方向看。一个社会不仅可能拥有知识却没有生产能力，也可能在知识仍然存在的时候，失去原来的生产能力。
美国曾经拥有庞大的制造业体系。随着国际分工深化，许多制造活动逐渐转移到海外。日本、韩国、中国等地的产业成长当然包含大量自主创新，不能简单理解为把美国的工厂搬走。但生产活动的地理转移，确实意味着技术人员、供应商、熟练工人和管理队伍在另一些地方重新积累。
假如今天美国希望恢复某些长期外移的制造环节，会遇到什么？产品设计可能仍在，工程知识并未消失，设备也能采购。专利还在美国公司手里，教科书还在美国大学的图书馆里。为什么不能立即恢复原有产能？因为稳定生产需要熟悉工艺的技术人员，需要能组织现场的管理者，需要懂得操作和质量控制的工人，还需要彼此配合的供应商。机器怎样布置，工序怎样衔接，异常怎样诊断，检验怎样安排，这些都要在实际生产中反复调整才能定下来。
长期不再从事某种生产活动，相关人员会退休、转行、迁移。供应商网络会改变，原来熟悉的协作关系会慢慢消失。于是出现一个很特别的现象：一种生产能力消失了，而没有任何一项科学知识失传。这里有一个耐人寻味的不对称。生产能力可以在几年里散掉，重建却要慢得多。散掉的时候没有人察觉，因为知识没有少一页。重建的时候才发现，少的东西不在纸上。
这当然不是说所有制造业回流都同样困难。它说的是，重建的成本不能只用设备和图纸来估计。
发展中国家面对的问题，是生产能力怎样形成。产业外移地区面对的问题，是生产能力怎样折损和重建。两者指向同一种有自己生命周期的生产能力。它的形成、折损和重建，都不与知识的存量同步。
三、图纸就在手里，为什么桥梁修不起来？
有一次，我向一家大型国有工程企业的负责人了解疫情期间的项目复工情况。这家企业承担过许多大型基础设施建设，有工程师，有图纸，有设备，有一套成熟的项目管理体系。可是有些工程迟迟不能复工。
我问为什么。他说，长期合作的农民施工队还没有从老家回来。我感到奇怪：这么大的企业，难道不能换一批人？他解释说，那些施工人员非常懂工程。图纸读得好，知道怎样把设计要求变成现场作业，知道怎样安排工序，遇到问题知道怎样处理。临时换人，很多事情要从头摸索。他感叹，那些农民工里的师傅看图纸的本领实在了不起。
我后来常想这件事。一家企业的生产能力，有一部分不在它的工资单上。那些施工队不是它的员工，却是它的一部分。企业能开工，靠的不只是自己拥有的东西，还有它和这群人之间多年形成的默契。这就是为什么，“换一批人”这个在账面上完全可行的方案，到了现场行不通。
这个故事里有两样不同的东西，要分开看。第一样是施工队自己的本领。设计人员掌握结构、材料和规范，施工队掌握现场工序与操作技巧，工长知道怎样协调不同工种。施工人员并不是在机械地执行指令。图纸上画的是设计意图，怎样在一片真实的场地上实现它，要靠他们的解释。从这个意义上说，他们也在参与设计。这种本领是他们自己的，走到哪里带到哪里。
第二样是他们和这家企业之间的东西。长期合作的工程师、工长和施工队，已经形成一套彼此熟悉的工作方法。谁负责什么，怎样提出问题，怎样确认修改，什么时候停工检查，遇到异常先找谁，各方早有共同预期。这套预期不写在任何文件里，也不归任何一方单独所有。换一批同样熟练的人来，他们各自都懂行，这套共同的工作方法仍然要重新建立。
停工也可能直接来自劳动力短缺，这个故事不能证明损失都来自协作关系。但它让我们看到一种可以单独拿出来看的能力：不同角色之间已经形成的共同校准。它并不完整地存在于工程师身上，也不完整地存在于施工队身上。它存在于他们之间。
四、增长理论把什么放进了生产率？
从这三个例子回到理论。它们提出的问题可以分成三层，每一层经济学都已经走了很远，本文想做的是顺着这条线再往前走一步。
第一层问题：知识怎样成为生产率？索洛把现实生产率的变化写进一个技术参数。这个参数后来被称为全要素生产率，实证研究用投入和产出的关系把它测出来。阿布拉莫维茨在1956年就提醒人们，这个残差是“我们无知的度量”，里面混着知识、组织、管理、资源配置，还有测量误差。阿罗1962年的“干中学”揭示了一个过程：生产本身产生知识。阿特金森和斯蒂格利茨接着指出，这种学习是局部的，你在哪种技术上生产，就在哪种技术上进步。再往后，技术采纳的研究追问企业为何不使用已知技术，吸收能力的研究问同样的外部知识为什么在不同企业产生不同效果，适宜技术的研究问先进国家的技术是不是适合后发国家的要素结构。科明和霍拜因等人的跨国技术扩散研究进一步提示我们：一项技术何时被引进，与它后来被使用得多广、多深，是两个不同的问题。这一层的结论可以归结为一句话：从知识传播到生产率实现，中间有复杂的条件。
第二层问题：组织为什么有生产性？科斯问企业为什么存在，威廉姆森问活动由市场还是由科层来组织。钱德勒写《看得见的手》，讲现代大企业的管理层级怎样取代市场协调。彭罗斯更早就说过，企业成长的真正约束是管理团队的扩张速度，新经理要在共同工作里学会这家企业的做法，这个过程不能靠雇人加速。布鲁姆和范里宁测量了几千家企业的管理实践，发现差异巨大、持久，而且能解释生产率的差别。这一层的结论是：组织不只是容纳生产要素的容器，它本身就是生产率的来源。
第三层问题：组织的知识存放在哪里？纳尔逊和温特在1982年提出组织惯例，说企业的生产知识不完全存放在个人头脑里，而在成员之间稳定、重复的互动模式里。团队专用人力资本的研究发现，一群人一起工作久了，生产率会高于把他们拆开再各自和别人组合。野中郁次郎讲企业怎样在互动里创造和转化知识。这一层的结论是：有一部分生产知识，不属于任何个人，而属于人与人之间的关系。
三层加起来，我们已经能看清一件事：现实生产率不仅来自知识本身，也来自组织知识的形成与运用。沿着这条路再往前走一步，就会碰到一个很具体的问题。一个组织解散了，原来共同形成的生产能力损失了什么？参与者各自带走了什么？这些人进入新的组织，为什么有时很快就能合作，有时仍然要漫长的磨合？惯例研究已经揭示组织具有集体知识。本文沿着这一认识，进一步区分两类东西：组织解散时失效的共同校准，与个人能够带走的角色经验，追问新组织怎样重新形成。彭罗斯强调管理团队的成长受到共同实践的制约。这里再把工人一侧的经验纳入分析，考察两侧经验的组合怎样影响校准时间。
所以本文要做的事可以说得很清楚。它不提出一套新的增长理论。它提出一个难以解释的增长现象，然后在增长理论里切开一个以前被过度压缩的环节，把里面的东西拿出来看。增长理论把从知识到产出的全部困难压进一个生产率参数。本文想把其中必须由不同角色共同实践才能形成的那一部分单独拿出来，给它一个名字，描述它的形成、折损和重建，说明它为什么值得作为一个独立的研究对象。
知道一家工厂应该归谁所有，不等于知道这家工厂第一天早晨该怎么开工。管理实践的研究常常考察已有企业怎样设定目标和考核，而建立第一家工厂还要解决更前面的困难：十个人来了，任务怎样分，怎样交接，谁核验谁的工作，出了差错怎样纠正。这是组织构成层面的问题。下面两节就从这里开始。
五、组织技术：把许多人变成一个生产单位
一个人知道怎样制造某种产品，不意味着他知道怎样组织十个人生产它。第一天早晨，十个人来到车间。谁负责什么？一个工头管几个人？哪些工作能够计件，怎样核对数量，返工怎么办？这些都有具体做法，而且做法有好坏之分。做法有好坏，正是技术这个词的本义。
我把这套做法叫作组织技术。它包括任务怎样分解，信息怎样传递，多人的行动怎样衔接，生产怎样核验，异常怎样处理。它既包括一个成熟的组织怎样运转，也包括一个组织最初怎样建立起来。
技术是做法，制度是权利与权力的安排。制度决定谁有权发令，谁承担责任，谁享有剩余。组织技术回答的是另一类问题：决定了谁发令以后，令怎样发，发到哪一级，怎样确认下面听懂了。两者相互依赖，却可以分开来看。同一种车间的组织方法可以在不同所有制下运转。同样的产权制度下，企业的组织能力也可以差别巨大。我们常常把后一种差别归因于文化、企业家或者运气，其实它有自己的内容，可以被描述，也可以被学习。
组织技术还有一个特殊的性质：它并不完整地归任何一个人所有。厂长知道怎样排班，工人也必须知道怎样按班次工作。线长知道怎样检查，工人也必须理解检验和返工的程序。一名经验丰富的厂长到一个从未办过工厂的村庄，未必能很快建起生产线。一群有经验的工人如果缺少懂得组织生产的人，也未必能自己建厂。
只说双方互补还不够。机器和劳动同样互补，经济学对互补的处理已经很熟练了。这里特殊的地方在于，不同参与者必须通过一起工作，才能形成彼此适配的行动规则。一个人可以独自学会开车床，却不能独自学会和别人交接。交接只能和别人一起学。组织技术中有些做法可以写成手册、程序和软件；另有一部分必须在具体参与者之间共同实践才能形成。这后一部分共同校准，只能在相应的协作关系中发挥作用。
第四节说过，纳尔逊和温特已经发现组织里有集体知识，团队专用人力资本的研究也已经发现，一起工作久了的人更有生产率。本文接着他们往前走一步，问这种集体知识是由什么构成的。哪一部分可以被个人带走？哪一部分只能留在关系里？组织解散的那一刻，两部分各自怎样变化？下一个组织形成的时候，前一部分怎样帮助后一部分更快生成？这几个问题连在一起，就从“组织有集体知识”这个一般判断，走到一套可以写下来的动态规律。这就是下一节。
六、双侧校准：组织能力怎样形成、消失与重建
设想一个管理者与一个工人团队开始合作。他们要共同确定生产节拍、岗位边界、质量标准、报告程序和异常处理方式。这些参数必须双方都理解，双方都能执行。任何一侧都不能单独定下来。管理者宣布一个节拍，工人做不到，这个节拍就不存在。工人提出一个节拍，管理者不认，它也不存在。
有经验的人未必预先知道新厂的最佳节拍。一个经验丰富的管理者，知道怎样和工人一起确定节拍：怎样试运行，怎样观察瓶颈，怎样调整流程，怎样确认对方已经理解。他不一定预先知道这家新厂的节拍该是多少。他从上一家厂带走的，是校准的程序。他留在上一家厂的，是那套已经完成的共同校准。这就是组织知识与普通技术知识的不同。经验是掌握校准程序，不是预先知道每一个正确答案。所以经验再多，校准也不能省掉，只能缩短。
双方都有相容的角色经验，仍然需要就新厂的具体参数达成一致，但往往能够较快完成。只有一方有经验，另一方需要被带着学。双方都没有经验，还要先摸索怎样一起试验和核验。三种组合对应三种校准时间。
一个极简的表达足够说明问题。设要校准 K 项参数，缺乏角色经验的一侧需要a 单位时间学习角色；双方都熟悉程序时每项校准时间为b₀，只有一侧熟悉时为b，两侧都不熟悉时为b′，其中0＜b₀＜b＜b′。三种校准时间分别为Kb₀、a＋Kb 和2a＋Kb′。这个表达式不是为了精确描述某家工厂。它只说一件事：新组织形成的时间，取决于双方经验的组合，而不是任何一方经验的多少。
这里必须区分两种存量。第一种是个人携带的角色经验。管理者和工人各自保留，可以带到别的组织去用。第二种是运转中的共同校准：参与者已经确定、理解并持续执行的一套具体参数。它存在于匹配之中，不能由任何一侧单独带走。舞步可以一个人练，两个人跳得合拍，是一起跳出来的。换一个舞伴，舞步还在，合拍要重来。
我们可以把一个地区运转中的共同校准记为Q。它取决于有多少匹配已经完成校准并正在运行。知道一个地区有多少有经验的管理者和有经验的工人，仍然读不出Q。这些人可能正在组成稳定的组织，也可能刚刚失业，分散在不同地方。一个地区也可能有大量由外来管理者主持的工厂，却没有多少本地的有经验的管理者。这是本文最想守住的一条：组织校准的存量，不是个人角色经验的加总。
组织解散时，共同校准不再作为运转中的整体存在。个人经验留下了。下一次匹配要重新校准。经验缩短重建时间，却不能替代校准。这正是组织能力与个人技能的差别，也是知识还在，生产能力却消失了这句话的准确含义。
共同校准的可携带程度，在不同情形下并不一样。原来的班组可能整体转到新企业去，原来的管理者和工人可能再次合作，标准化的流程可能已经写进文件、软件和设备的设置里。在这些情形下，旧的校准有一部分跟着走。所以匹配专用性是一个程度变量。这恰好能解释为什么有的组织很容易复制，有的很难：复制的难易，取决于校准的多少能被写下来，多少只能在人与人之间重做。
把这种东西叫组织资本并非不可以。但资本这个词容易让人以为它可以像设备一样整体搬走，或者像专利一样整体转让。我在这里用共同校准，是想守住一个区分：匹配中的约定，和参与者可以携带的经验，不是一回事。
七、为什么第二家工厂更容易？
第一家工厂生产产品，同时也在培养下一家工厂需要的人。工人学会按班次工作、接受检验、同其他岗位配合。一部分人成为班组长、车间主任和厂长。人离开企业以后，带着不同角色的经验进入新的匹配。
马歇尔讲产业聚集的时候说，行业的秘密在空气里。克莱珀研究底特律的汽车业和硅谷的半导体业，发现成功企业的衍生企业是产业集群最重要的来源，衍生企业带走的恰恰是在母公司里学到的组织做法。本文想把这两位先贤的发现说得更具体一点：秘密不在空气里，在曾经坐过各个岗位的人身上，在他们彼此熟悉的工作方法里。一个地区的产业历史积累的，不只是专业技能，还有能够在新组织里重新组合的角色经验。这是一种新的集聚机制。已有工厂不只提供供应商、客户和熟练工人，还培养未来建立生产组织所需要的各种角色。新组织能够更快校准，地区复制组织的能力随之增强。所以第一家工厂最难，第二家容易。
这种积累有一个条件：本地人必须有机会担任相应的角色。如果外资厂长期由外来团队管理，本地人主要从事生产，当地积累的就是工人侧的经验。工厂可以经营得很好，产量可以不断增加，本地的管理经验却未必增加。成功生产不自动意味着获得独立复制同类组织的能力。
这是全文最反直觉的一个结论。一个地区的工厂可以越来越多，产出越来越高，效率越来越好，而这个地区独立创建新组织的能力可以一直不变。生产活动本身不能复制生产活动。能复制生产活动的，是曾经在生产活动里担任过各种角色的人。阿罗的“干中学”说，做得多就学得多。本文要补一句：学到什么，取决于你在生产里做的是什么。一个地区从生产里学到了什么，取决于本地人在生产过程里站的是哪个位置。
这与工资激励无关，也不依赖外来管理昂贵的假设。即使外来管理免费提供，本地人没有坐过管理岗位，就不能通过管理实践积累那个角色的经验。一个地方学会了生产，未必已经学会组织生产。
共同校准的想法还给出一个关于重建的预测。两个地区原有的工厂一样多，经验人员的总数一样多。一个地区的经验分布在管理者和工人两侧，另一个地区的经验集中在工人侧。假如原有组织同时解散，外来管理者离开，前者可以较快重建双方都有经验的匹配，后者必须让新手管理者先学角色，再重新校准。差距的一部分是普通的角色学习，另一部分是双方共同校准的额外时间。
这就是本文所说的历史依赖。通常讲技术扩散，历史决定一个地区拥有多少知识。在这里，历史还决定知识分布在哪些角色的人身上，以及哪些人曾经一起工作过。同样数量的经验人员，可以有完全不同的组织重建能力。这比一般说的路径依赖要具体得多，也可以测量。
因此，生产组织可以迅速消失，而重建需要时间。过去谁坐过哪个岗位，决定了冲击之后恢复的速度。这也解释了为什么产业外移的损失常常被低估。在统计上，知识没有少，熟练工人至少头几年也还在。少的是匹配，而匹配不进入任何统计。
八、苏南、珠三角、温州与东北：不同的组织经验历史
中国改革开放以后的区域工业化，可以用这个角度重新看一遍。四个地区的历史组织形式不同，积累的组织经验也不同。
苏南在改革开放以前就有社队企业和集体工业的连续实践。许多农村劳动者、基层管理者和地方干部，并不是第一次接触工厂组织。工人侧的经验有，基层管理侧的经验也有，地方干部还有组织资源和调解纠纷的经验。这种存量可能降低后来乡镇企业、改制企业和私人企业的建立成本。值得追踪的，是哪些组织岗位连续存在，哪些人从旧组织转入了新组织。
珠三角走的是另一条路。港澳台商人带来订单、设备和管理团队，当地劳动者进入现代工厂。工厂建了多少，是次要的问题。本地人什么时候成为班组长、车间主任和厂长，才是关键。只有实际担任这些岗位，管理侧的组织经验才会在本地积累。珠三角后来出现了大批本地企业。管理岗位的本地化是否以及在多大程度上推动了这一变化，值得追踪。开放得早的地方和开放得晚的地方，后来的样子应该不一样。
温州早期大量利用家庭作坊和亲属网络。亲属之间已有信任和分工习惯，使小规模生产能够较快启动。但亲属型组织和陌生人协作组织需要的是两种经验。长期只与亲属合作，未必培养出管理几百名陌生工人按流水线协作的能力。什么时候开始大量雇用非亲属，什么时候引入有经验的线长，是组织方式升级的关键节点。
东北提醒我们另一件事：工业经验有组织方式上的专用性，不能简单加总。那里拥有庞大的产业工人队伍和国有企业的管理人员，却没有因此自动形成同等规模的民营工业。国有企业里的角色经验与竞争性私营企业的组织方式，可能只有部分相容。产权、市场环境、企业家供给和地方政策也都影响结果。值得研究的是，下岗工人到南方的私营厂以后获得了哪些岗位，是否获得了工资溢价，不同岗位的经验能不能跨所有制使用。
四个地区合起来，提出一个新的看历史的角度：一个地区今天能够建立什么组织，与它过去曾经怎样组织生产有关。历史上的生产活动，通过人的角色经验延续很多年。
九、福建的房地产村：组织技术怎样成片传播
我有一位学生来自福建一个村庄。据他介绍，村里约五百户，其中两百户参与房地产开发，在全国成立了一千多家房地产公司。这些数字我没有核实过，但这个村庄怎样一步一步学会盖房子，是一个值得讲完整的故事。
上世纪八十年代，村里分地以后人均土地很少，有人组织施工队到县城承接建设工程。早期从承接建设工程做起。施工队里有人会看图，有人会带班，有人会跟甲方打交道。后来遇到工程款不能及时支付的情况，有的施工队拿到一块地作为补偿。地在手里，总不能空着，于是自己盖房、自己卖。从施工承包到房地产开发，组织要完成的任务复杂了许多。原来只要把楼盖起来，现在要买地、报批、融资、设计、施工、销售，每一样都是新的。参与者边做边学。
一个人做成以后，邻居跟着学。有人掌握施工，有人熟悉项目管理，有人懂销售、资金和土地事务。项目班子在不同城市之间流动。一个项目做完，班子可能解散，有人留在原地，有人去别的城市，各自与新的伙伴组成项目团队。几十年下来，村里人参与的建设项目遍及多地。
我曾问他，房地产开发这么复杂，你是怎么学会的？他说，邻居都会，慢慢自己也会了。这句话里有一整套经济学。邻居都会说的是，这种能力的持有者不是某一家公司，而是这个村庄。公司成立又解散，项目开始又结束，村子里的经验不散。慢慢自己也会了说的是，这里传播的既有技术知识，也有角色经验和共同校准的方法。一个项目结束，原来具体的协作关系未必完整保留，但参与者知道怎样在新项目里较快重新分工、核验和配合。他们学会的，是校准的程序。
房地产开发当然还依赖土地机会、融资关系和项目信息，这些村里人也是共享的。要分清哪一部分来自组织校准，将来要去问三件事：项目班子怎样流动，不同的人分别在哪些岗位上学会了什么，一个新项目从筹备到稳定运行要多久。但这个故事已经说明了一件事：一个地区的组织经验，可以通过同乡、亲属和职业网络成片积累，并不总要依靠大型企业或者产业园区。第一批项目支付了学习成本，后来者使用前人留下的角色经验。这种能力附在一群人身上，而不附在任何一家企业身上。
十、出海：为什么成熟的企业到了海外还要重新学会生产？
我这几年一直在讲一个判断：未来二十年，中国可能有大量企业到海外扎根，规模或许达到十万家以上。不是出口，是在别人的土地上建厂、雇人、组织生产。这可能成为规模可观的跨国生产组织迁移，令人联想到上世纪八十年代港澳台资本进入珠三角以及日本企业在北美设厂的经历。本文讲的第一家工厂为什么难，对这些准备出海的企业来说不是学术问题，是它们明天早晨就要面对的事。
先问一个看起来奇怪的问题。一家在中国已经成熟运转了二十年的企业，到了越南或者墨西哥，为什么还要重新学会怎样生产？它的设备、图纸、工艺参数、供应商清单，都能带去。它的厂长、车间主任、线长、设备和质量的骨干，也能带去。它甚至可以把整个班组带去。它带不去的，是当地的工人侧经验、当地的供应商和当地的工作环境。而生产能力不是知识的清单，是这些东西和它带去的东西校准以后的结果。到了海外，校准要重做一遍。
而且海外的校准比国内多。当地工人也许有经验，也许没有，有也未必是和中国管理者相容的那一种。语言、工时习惯、工会、宗教节日、法律对加班和解雇的规定、当地供应商的交付节奏，都是国内从来不用校准的参数。用第六节的符号说，海外的K 更大，两侧经验相容的程度更低，所以a 和b 都更长。我想把这件事叫作组织校准距离：两个地区之间生产组织的相容程度。它不是地理距离，也不是文化距离。墨西哥和中国的地理距离远，但一家在东莞做过电子代工的企业，到墨西哥北部的电子产业带去，组织校准距离可能不大，因为那里的工人侧经验已经被美国和亚洲的工厂培养过。反过来，到一个从未有过现代工厂的地方去，地理再近，组织校准距离也远。这个概念现在还不是一个严格的变量。它是一个值得研究的角度：决定组织技术能不能传过去、多快传过去的，可能是组织校准距离，而不是地图上的距离。
这里有一个重要的先例可以看。1984年，丰田和通用在加利福尼亚的弗里蒙特合办了一家工厂。那座工厂原来是通用的，以质量差和劳资冲突出名，关掉了。合资企业重新聘用了原厂约85%的工会工人，采用丰田的生产组织方式。到1986年，其生产率已超过通用旗下其他工厂，质量也显著改善。这个例子常被用来说明管理方法的重要。我从本文的角度看，它说的是另一件事：那些工人不是被替换掉的，是被重新校准的。丰田带去的是管理侧的程序，工人带来的是工人侧的经验，两侧在新的参数上重新对齐。校准花了时间，但远少于从零培养一批工人。日本企业八十年代在北美的成功，很大程度上是这种双侧校准的成功。
对出海企业来说，接下来的问题是本地管理人员的培养。第七节的逻辑在这里原样适用。企业可以把中国的线长和班组整个带出去，核心工段由中国人运转，当地人在外围。整支班组迁移能够保存原有成员之间的一部分协作规则，通常有利于缩短初期磨合。但国内运转中的共同校准Q 不能直接搬到海外：新厂仍要同当地设备、供应商、人员和工作环境重新校准。如果当地人长期只在生产岗位参与，那么本地积累的主要是工人侧经验；后续工厂仍可能依赖外来管理人员。但只要当地员工逐步进入线长、车间主任等岗位，整队迁移也可以成为培养本地管理经验的起点。企业也可以从第一天起就让当地人进入班组长、工段长的岗位，中国管理者带着他们学。这可能增加初期培训与协调工作，也可能利用当地人员对语言、劳动市场和制度环境的熟悉而缩短某些校准。几年以后，本地管理经验能否积累，取决于当地人员实际担任了哪些岗位、持续参与了多久。真正值得比较的，是不同人员配置如何同时影响当前校准速度与未来组织复制能力。
如果海外设厂规模持续扩大，完全依靠外派管理团队可能受到人员供给、成本与当地用工环境的限制。珠三角的历史在这里会倒过来重演：当年港台企业什么时候让本地人当厂长，中国企业现在就要回答同样的问题，只是这一次站在港台企业的位置上。本地管理岗位怎样逐步形成，可能影响多年以后当地能否自主建立更多同类组织。这是出海企业值得提前思考的问题。
还有两件事值得出海企业从本文里带走。一件是第三节的施工队。企业在国内的生产能力，有一部分从来不在它的工资单上，在它多年合作的供应商、施工队和外协厂里。到了海外，这些人不在。企业会发现自己比在国内小了一圈，很多原来不用操心的事忽然要自己做。重建这一圈协作关系，有时可能比重建车间本身更费时。另一件是第九节的村庄。第一批出海的企业支付了校准的学习成本，后来的企业可以用它们留下的人。所以出海企业在同
本文由 AI图鉴自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
Chen Yuyu, Professor, Guanghua School of Management, Beijing University, Director, Economic Policy Institute, Beijing University
Developed countries have long been developing countries' artificial intelligence. The world knows how cars are built, how power plants are built, but there are a few countries that have really come after them for 100 years. Today AI made knowledge cheaper, so the old question came back: why did cheap knowledge not automatically turn into growth?
This article opens a long-term condensed link in the growth theory: how a group of people can become a production organization that can share knowledge. The authors present the concepts of “organizational technology” and “bilateral calibration”, which explain why production capacity can disappear with a lot of knowledge, why the first plant is the hardest, the second is easy, and why a region learns production without necessarily learning to organize production. This perspective was used to focus on the industrialization of the Sunan, Pearl Triangle, Wenzhou, North-East, how a village in Fujian could become a piece of learned real estate, and two things that were happening: the large-scale outing of Chinese enterprises, and where new organizations came from after the contraction of manufacturing.
The article ended up back in AI. According to the authors, the contribution of AI to growth depends on whether it changes the availability of knowledge, organizational calibration, or both; the same technological potential is that both fast-to-school and slow-to-calibration industries move out of two completely different productivity paths.
It's a long thought, not a short comment. It raises questions, synthesizes knowledge, gives a new subject and clearly delineates its own interpretation boundaries. A reader who would like to take 40 minutes to read it.
Why is it that the world already possesses a great deal of advanced knowledge and that many countries have been unable to achieve rapid growth? Developed countries have long been developing countries' “artificial intelligence”, and AI today makes knowledge creation and acquisition easier. However, knowing how to produce is two different processes, as is actually developing productive capacity.
This paper combines the understanding of growth economics, technology diffusion and organizational economics and proposes a fresh look at the transformation of knowledge to productivity from the perspective of organizational technology. Among them, the “two-sided calibration” describes the work rules that different actors have developed to suit each other through common practice. Individuals can take their role experience, and the common calibration that has been developed cannot be carried by either side alone. This gives production organizations unique patterns of formation, accumulation, loss and reconstruction.
This perspective has contributed to a renewed understanding of industrial clustering, regional industrialization in China, enterprise access to the sea and industrial structural transformation, and has also provided new insights into industry differences that explain the growth effects of AI. AI may either increase the availability of knowledge or shorten or even change the organizational calibration process. However, real growth also depends on the reallocation of resources and the discovery of consumption needs.
The replicability of production knowledge and the replicability of production organizations are of two different technical nature. Understanding how the two are connected is an important direction in studying economic growth in the age of artificial intelligence.
Know how to do one thing, and a group of people make it, and there's a distance.
1. Developed countries have long been developing countries ' artificial intelligence
Today, there are high expectations of artificial intelligence driving economic growth. Machines are able to program, design products, explain technical difficulties and also begin to participate in scientific discoveries. Will the cost of knowledge production be significantly reduced, with which economic growth rates will increase significantly?
The problem is not new. Before answering it, let us look at an experiment that has been going on for more than a century. For developing countries, developed countries have long been their artificial intelligence. The world knows how cars are made, how power plants are built, how medicines are produced and how modern plants are managed. Equipment could be purchased, engineers could be hired, students could be sent out for study and multinational enterprises could even move the entire production line. Knowledge is not free. Intellectual property rights, trade restrictions and the political environment can pose obstacles. However, the cost of acquiring readily available knowledge is much lower than the cost of rediscovering it.
It is assumed that in 1960 a person had taken the reasoning that the world’s production knowledge was almost readily available to Ghana and India, that advanced countries had already paid the cost of discovery and that those countries would need only to use it later, so they would soon catch up. This reasoning was not absurd at the time. Gshenkron says "the advantage of backwardness" and that's it. Many development economists generally share the same view. Sixty years have passed, and we all know. The catching up of economies such as Japan, Korea and China is remarkable, precisely because they are a minority. A large number of economies have long been exposed to advanced knowledge but have not developed commensurate productive capacities. This has been repeatedly confirmed in the literature on development accounting: the difference in per capita income between countries is largely not explained by material capital and years of education, while the remaining part is classified as “full factor productivity”.
And a more extreme ideological experiment. If today the world ' s most advanced production knowledge is handed over in its entirety to a country with very low productivity, so are technological drawings, production processes, management manuals, and all scientific knowledge. Could it become a developed economy within 10 years? Nobody's gonna bet like that. So, what is it missing? Capital, institutions, markets and infrastructure are certainly lacking. But there is another category that is not on any list of knowledge and that cannot be obtained by delivery. It can only grow slowly in actual production activities. It is the ability to integrate knowledge, people and resources into real-life productive activities.
This is what this paper is about. We have studied deeply how knowledge is created and disseminated. We do not know enough how a society acquires the capacity to use knowledge. The availability of knowledge and its economic realization are two things. Taking both as one is the whole mistake of the 1960 reasoning. Many of today's reasoning about AI is repeating this mistake. By just how much knowledge the machine can do, it can figure out how much GDP can grow, skipping the hardest step in the middle.
AI made this old problem more acute. Knowledge has become cheaper, and it is now becoming cheaper. If cheap knowledge has not been automatically translated into growth in the past, how can cheaper knowledge be?
One more floor. Growth is not the end of what is known. Higher incomes, changing demand, shrinking old industries and growing new ones. When a group learns to do one thing, society constantly lets another learn to do another. This article begins with the first half, and how the first factory was built. After that, when the old factory was closed, where did the new organization come from?
Knowledge remains and productive capacity may disappear
It can be seen in the opposite direction. A society may not only possess knowledge without productive capacity, but may also lose its original productive capacity while knowledge remains.
The United States had a huge manufacturing system. As the international division of labour deepened, many manufacturing activities gradually shifted abroad. Industrial growth in places such as Japan, Korea and China certainly involves a great deal of autonomous innovation, which cannot simply be understood as removing American factories. However, the geographical transfer of productive activities does mean that technicians, suppliers, skilled workers and management teams have re-emerged in other places.
What would the United States encounter today if it wished to restore some of the manufacturing links of long-term emigration? Product design may still be in place, engineering knowledge is not lost and equipment can be procured. Patents are also in the hands of American companies and textbooks are also in libraries at American universities. Why not immediately restore the original capacity? Since stabilization of production requires skilled people who are familiar with the process, managers who can organize the site, workers who know how to operate and quality control, and suppliers who work with each other. How the machines are programmed, how the processes are connected, how the diagnosis is unusual, and how the arrangements are tested, will have to be adjusted over and over again in actual production.
The person concerned will retire, transfer or relocate from a productive activity for a long period of time. The supplier network will change and the previously familiar collaboration will slowly disappear. A particular phenomenon follows: a productive capacity has disappeared without any scientific knowledge being lost. There's an interesting asymmetry here. Productive capacity can be dispersed over the years, but reconstruction is much slower. No one noticed it when it was scattered, because there was not a single page of knowledge missing. When we rebuild, we find that less is on paper.
This, of course, does not mean that all manufacturing returns are equally difficult. What it says is that the cost of reconstruction cannot be estimated solely by equipment and drawings.
The question for developing countries is how productive capacities are formed. The problem with out-migration is how productive capacities can be eroded and rebuilt. Both point to the same productive capacity with its own life cycle. Its formation, loss and reconstruction are not synchronized with the stock of knowledge.
iii. The drawings are in hand. Why can't the bridge be repaired?
On one occasion, I asked the head of a large State-owned engineering enterprise about the resumption of work during the epidemic. The enterprise has been responsible for a number of large infrastructure, engineers, drawings, equipment and a mature project management system. However, some work was delayed.
I asked why. He said that the long-standing cooperative campesino construction team had not returned from its home. I was wondering: can't there be another group of people in this big business? He explained that the construction staff were very well-versed in engineering. The drawings are well read, know how to translate design requirements into on-site work, know how to organize the process, and know how to handle problems. There's a lot to be done. He lamented the fact that the masters of the peasant workers had great skills in reading the drawings.
I thought about it a lot later. Part of an enterprise ' s productive capacity is not on its payroll. The construction teams are not its employees, but part of it. Businesses can start working not only on what they own, but also on the understandings that they have developed over the years with this group of people. That is why the “replacement of people” -- an entirely viable book-to-book option -- does not work on the ground.
There's two different things in this story. The first is the construction team's own skill. Designers master structures, materials and specifications, construction teams master on-site process and operating skills, and the Chief of Staff knows how to coordinate different types of work. The construction personnel did not carry out the instructions mechanically. The drawings are designed to achieve it on a real site, depending on their interpretation. In that sense, they are also involved in design. That's where they come from.
The second is between them and this business. Long-standing cooperative engineers, chief engineers and construction teams have developed a familiar working methodology. The parties had a common vision of who was responsible for what, how to ask questions, how to confirm the changes, when to stop inspections, who to look for in the first place. It is not written in any document or owned by any party alone. The common working method still needs to be re-established with the arrival of a new team of equally skilled people who know each other.
Stoppages may also come directly from a shortage of labour, a story that does not prove that the losses were all due to collaborative relationships. But it shows us a capability that can be seen alone: a common calibration that has developed between different players. It does not exist in complete on the engineers or on the construction team. It exists between them.
What does the growth theory put into productivity?
From these three examples back to theory. Their questions can be divided into three layers, each of which has gone a long way, and what they want to do here is to follow this line.
First tier: How can knowledge become productivity? Soro has written changes in real productivity into a technical parameter. This parameter was later referred to as total factor productivity, measured by the correlation between inputs and outputs of empirical research. Abramowitz was a reminder in 1956 that the disability was a “measure of our ignorance”, mixed with knowledge, organization, management, resource allocation and measurement errors. Aro's “dry secondary school” in 1962 revealed a process: production itself produces knowledge. Atkinson and Stiglitz went on to point out that this learning is partial and that the technology in which you produce is the technological progress in which. In the future, research on technology adoption asks why firms do not use known technologies, research on absorptive capacity asks why the same external knowledge has different effects in different firms, and research on appropriate technologies asks whether technology in advanced countries is suitable for the factor structure of the latter. Cross-border technology diffusion studies by Komin and Hobain and others further point to two different issues: when a technology is introduced, and how widely and profoundly it is subsequently used. The conclusion of this layer can be summed up in one sentence: from knowledge diffusion to productivity achievement, there are complex conditions.
Second tier: Why is the organization productive? Cos asked why the business existed, and Williamson asked whether the activities were organized by the market or by science. Chandler wrote Visible Hands on how the management hierarchy of modern large enterprises could replace market coordination. As Penrose said earlier, the real constraint on corporate growth is the speed at which the management team expands and the new manager learns how to do business together, a process that cannot be accelerated by hiring. Bloom and Van Reyning measured the management practices of thousands of enterprises and found that the differences were large, persistent and explained the differences in productivity. The conclusion of this layer is that the organization is not just a container that accommodates factors of production, but is itself a source of productivity.
Third tier: where is the knowledge of the organization? Nelson and Winter introduced organizational practice in 1982, stating that the production knowledge of enterprises was not entirely in the individual mind, but in stable and repetitive modes of interaction among members. Team-specific human capital studies have found that a group of people has worked together for a long time and that productivity is higher than breaking them apart and combining them with others. Nobujiro said how enterprises create and transform knowledge in their interactions. The conclusion of this hierarchy is that there is a part of the knowledge produced that does not belong to any individual, but rather to human relations.
On three levels, we can already see one thing: real productivity comes not only from knowledge itself, but also from the formation and application of organizational knowledge. One step further along this road leads to a very specific problem. What was lost in the co-generation of productive capacity when an organization was disbanded? What have the participants taken? Why are these people entering the new organization, sometimes quickly, and sometimes still long-lasting? Practice research has revealed the collective knowledge of the organization. Following this understanding, the paper further distinguishes between two categories: common calibrations that fail when an organization is dissolved, and role experiences that individuals can take away, asking how a new organization can be re-established. Penrose stressed that the growth of management teams was constrained by common practices. Here, the experience of the worker side is included in the analysis, examining how the combination of experience on both sides affects calibration time.
So what is to be done here is clear. It does not propose a new set of growth theories. It presents an inexplicable growth phenomenon, and then cuts a link in the growth theory that was previously over-compressed and shows what is inside. The growth theory places all the difficulties from knowledge to output in a productivity parameter. It is the intention of this paper to single out those parts of it that must be developed by the common practice of different actors, giving it a name that describes its formation, impairment and reconstruction and explains why it deserves to be the subject of an independent study.
Knowing who owns a factory does not mean knowing how to start work on the first morning. The study of management practices has often looked at how enterprises have set targets and benchmarks, while the establishment of a first plant has to address the difficulties ahead: Ten people came, how the mission was divided, how the transfer was made, who checked whose work and how the error was corrected. This is an organizational dimension. The next two sections start here.
Organizational technology: transforming many people into a production unit
A person knows how to make a product does not mean that he knows how to organize 10 people to produce it. The first morning, ten people came to the workshop. Who's in charge of what? How many are there? Which jobs can be counted, how can the numbers be checked and what can be done to return to work? These practices are both concrete and mixed. There are good and bad practices, which are the very essence of the term technology.
I call it organizational technology. It includes how tasks are broken up, how information is delivered, how many people's actions are connected, how production is verified and how unusual. It encompasses both how a mature organization functions and how an organization was initially established.
Technology is the approach, and the system is the arrangement of rights and powers. The system determines who has the right to make an order and who is responsible and who has the residual. The organizational techniques answer another category of questions: determining who will issue the order, how, at what level and how to ascertain that it is understood. The two are interdependent but can be seen separately. The same workshop is organized in a way that can operate under different ownership systems. Under the same property rights regime, the organizational capacity of enterprises can vary significantly. We often attribute the latter difference to culture, entrepreneurs or luck, but it has its own content, can be described or learned.
Organizational technology is also of a special nature: it is not entirely owned by anyone. The director knows how to schedule and the workers must know how to work on a shift basis. The chief knows how to check, and the worker must understand the procedures for testing and returning to work. An experienced plant director to a village that has never operated a factory may not be able to build production lines quickly. A group of experienced workers may not be able to build their own plants if they lack people who know how to organize production.
It is not enough to say that the two sides complement each other. Machines and labour are equally complementary, and economics is already very skilled in dealing with complementarities. What is particularly special here is that it is through working together that the different actors must work together in order to develop compatible rules of action. A person can learn to drive a bed alone, but not to interact with others. You can only learn with someone else. Some of the practices in organizational techniques can be written into manuals, procedures and software; others must be shared among the specific participants to be developed. This latter part of the common calibration can only function in the corresponding collaborative relationship.
Section IV states that Nelson and Winter have found collective knowledge in the organization and that research on team-specific human capital has found that people who have worked together for a long time are more productive. This paper goes on to ask what this collective knowledge is made up. Which part can be taken by individuals? Which part of the relationship only stays? How did the two parts change when Division was disbanded? How does the first part help to produce the second part faster when the next one is formed? These problems are linked from the general judgement that “organization has collective knowledge” to a dynamic pattern that can be written down. This is the next section.
VI. Double-sided calibration: how organizational capacity is formed, disappeared and rebuilt
It is envisaged that a manager will start working with a team of workers. They jointly determine production rhythms, job boundaries, quality standards, reporting procedures and unusual handling. These parameters must be understood by both sides and implemented by both sides. Neither side can be left alone. The manager announced a beat, which workers could not do, which would not exist. The worker proposed a beat, which was not recognized by the manager and which did not exist.
Experienced people do not necessarily know in advance the best beats of the new plant. An experienced manager knows how to determine the beat with the workers: how to test, how to observe bottlenecks, how to adjust the process, how to confirm that the other side understands. He does not necessarily know in advance how many beats the new factory should have. He was taken from the previous factory in a calibration procedure. He stayed at the previous plant with the completed common calibration. This is the difference between organizational and general technical knowledge. Experience is mastery of the calibration process, not advance knowledge of every correct answer. So with more experience, calibration cannot be saved, but only shortened.
Both sides have compatible role experiences and still need to agree on specific parameters for new plants, but often they can be completed relatively quickly. Only one has experience and the other needs to be brought to school. Both sides had no experience and had to find out how to test and verify together. The three combinations correspond to three calibration times.
A very brief statement is sufficient to illustrate the problem. For the K parameter to be calibrated, a unit time is required on one side that lacks role experience to learn the role; each calibrated time is b0, only one side is b and neither side is b, of which 0<b<b<b>. The three calibration times are Kb0, a+Kb and 2a+Kb. This expression is not intended to describe an exact factory. It says only one thing: the timing of the formation of the new organization depends on a combination of experiences on both sides, not on the level of experience on either side.
A distinction must be made between the two stocks. The first is personal role experience. Managers and workers retain each other and can be taken to other organizations. The second is a functioning common calibration: a set of specific parameters that participants have identified, understood and implemented on an ongoing basis. It exists in a match and cannot be taken by either side alone. The steps can be practiced alone, and the two of you can dance together. We're gonna change the dance. We're gonna start again.
We can put the common calibration of a region into Q. It depends on how many matches have been calibrated and are running. Know how many experienced managers and experienced workers in a region can still not read Q. They may be forming stable organizations or may be newly unemployed and dispersed in different places. A region may also have a large number of factories run by external regulators, with few local experienced managers. This is one of the things that this paper wants most: the stock of organizational calibration, not the sum of personal role experience.
When the organization is dissolved, the common calibration ceases to exist as a functioning whole. Personal experience is left behind. Next match is to be calibrated. Experience reduces reconstruction time, but is not a substitute for calibration. This is precisely the difference between organizational capacity and individual skills, as well as the fact that knowledge exists and productive capacity disappears from the exact meaning of the phrase.
The portability of the common calibration is different in different contexts. The original group may be transferred to the new enterprise as a whole, the original manager and the worker may be working together again and the standardized process may have been written into the document, software and equipment set-up. In these cases, part of the old calibration followed. So matching is a level variable. This explains why some organizations are easy to replicate and some are difficult: the ease of reproduction depends on how much calibration can be written down and how much can be done only between people.
It is not impossible to call such things organizational capital. But the term capital can easily be assumed to move as a whole as an equipment, or as a patent. I'm here with a common calibration to keep the distinction: the agreement in the match, and the experience the participants can carry, is not the same thing.
Why is the second factory easier?
The first factory produces products and is also training those who need them in the next factory. Workers ' associations work in shifts, undergo tests and cooperate with other jobs. Some of them became class leaders, workshop directors and factory directors. When people leave the business, they enter new matches with experience from different players.
Marshall said when the industry gathered, it was in the air. Clerper studied the car industry in Detroit and the semiconductor industry in Silicon Valley, and found that derivatives from successful enterprises were the most important source of industrial clusters, and that derivatives took away organizational practices learned in parent companies. The paper would like to make the discovery of these two pioneers more specific: secrets are not in the air, they are in the hands of people who have served in various positions, and they are familiar with each other's working methods. The industrial history of a region has accumulated not only professional skills but also role experiences that can be regrouped in new organizations. This is a new gathering mechanism. Existing factories provide not only suppliers, customers and skilled workers, but also develop the various roles that will be required to establish productive organizations in the future. The ability of new organizations to calibrate faster increases the capacity of regional replication organizations. So the first factory is the hardest, the second is easy.
This accumulation is conditioned by the fact that locals must have the opportunity to assume the corresponding role. If foreign firms are managed by external teams for a long time, locals are mainly engaged in production, and what is accumulated locally is worker-side experience. Plants can operate well, production can increase, but local management experience does not necessarily increase. Successful production does not automatically mean acquiring the ability to replicate the same organization independently.
This is one of the most intuitive conclusions of the whole text. The ability of a region to create new organizations independently can remain the same. Productive activities themselves cannot replicate them. The reproduction of production activities is done by persons who have played various roles in production activities. Aro's “dry high school” says that much more is done. And here we have to add: what you learn depends on what you do in production. What an area has learned from production depends on where the natives are in the production process.
This has nothing to do with wage incentives and does not depend on costly external management assumptions. Even when external management is provided free of charge, locals do not hold managerial positions, and the experience of that role cannot be accumulated through management practices. A place has learned to produce, not necessarily to organize production.
The idea of common calibration also gives a projection of reconstruction. The two regions had the same number of former factories and the same number of experienced personnel. The experience of one region is divided between managers and workers, while that of the other is concentrated on the worker side. If the former organization were to be dissolved at the same time, external managers would leave, the former would be able to re-establish relatively quickly a match of experience between the two parties, while the latter would have to allow beginner managers to learn their role before recalibrating. Part of the gap is general role learning, and the other part is additional time for joint calibration.
This is what this paper calls historical dependence. Usually, technology spreads, and history determines how much knowledge a region has. Here, history also determines which players are spread knowledge and who have worked together. The same number of experienced persons can have completely different organizational capacity for reconstruction. This is much more specific and measurable than the generally stated path.
Thus, production organizations can disappear quickly and reconstruction takes time. Who used to be in what position determined the rate of recovery after the shock. This explains why losses from out-migration are often underestimated. Statistically, there is no shortage of knowledge and skilled workers are still present for at least the first few years. Less matches, but matches do not enter any statistics.
VIII. SUNAN, JUNG, WUNCH AND NORTH-NORTH: A DIVISION OF ORGANIZATIONAL EXPERIENCE
The regional industrialization of China after reform and opening up could be revisited from that perspective. The four regions have different historical forms of organization and have different organizational experience.
Before the reform and opening up of the country, there was a continuum of social enterprise and collective industry. Many rural workers, grass-roots managers and local cadres are not first contact with factory organizations. Worker side experiences, grass-roots management side experiences, local cadres, organizational resources and dispute mediation experience. This stock may reduce the cost of setting up later township enterprises, remodelling enterprises and private enterprises. It is worth tracking which organizational posts continue to exist and which individuals have moved from old to new organizations.
The Pearl Triangle goes another way. Hong Kong/Australia traders bring orders, equipment and management teams and local workers enter modern factories. How much has been built is a minor problem. It is crucial that locals become class leaders, workshop directors and plant directors. Only when these positions are actually occupied will management-side organizational experience accumulate locally. A large number of local enterprises subsequently emerged in the Pearl Triangle. Whether and to what extent the localization of managerial positions has contributed to this change is worth tracking. Early opening and late opening should be different.
Early in Winzhou, there was extensive use of family workshops and family networks. There is a tradition of trust and division of labour among relatives, allowing small-scale production to start faster. However, two types of experience are needed in relation to kin-based organizations and organisations working with strangers. Long-term cooperation with relatives alone does not necessarily create the capacity to manage the flow of water to hundreds of strangers. When large numbers of non-relatives are hired and experienced line lengths are introduced are key points for organizational upgrading.
The Northeast reminds us of another thing: industrial experience is organized in a way that is unique and cannot be simply summarised. There is a large industrial workforce and managers of state-owned enterprises, but this does not automatically result in private industry of the same size. Role experience in State-owned enterprises may be only partially compatible with the way competitive private enterprises are organized. Property rights, the market environment, the supply of entrepreneurs and local policies also influence results. It is worth examining what jobs laid-off workers have obtained after they went to private factories in the south, whether they have received wage premiums, and whether experience with different jobs can be used across ownership systems.
Together, the four regions offer a new historical perspective: what an area can build today is related to how it has organized production in the past. Historically, productive activities have lasted many years through human role experience.
IX. Real estate villages in Fujian: how organizational technology can be disseminated in one piece
I have a student from Fujian village. According to him, 500 families in the village Rio, 200 of them involved in real estate development, had established more than 1,000 real estate companies throughout the country. I have not verified these figures, but how the village learned to build a house is a complete story worth telling.
In the 1980s, since the division of the village, there was little land per capita and construction teams were organized to take over construction work in the county towns. The early stages begin with the commissioning of construction works. There are people in the construction team who read the drawings, people who lead the shift and people who deal with Alpha. Later, in the event that the works were not paid in time, the construction team received a piece of land as compensation. The land in hand can't be empty, so build and sell. From construction contracts to real estate development, the tasks to be performed by the organization are much more complex. As long as the building was built, it was now necessary to buy land, approve, finance, design, construct and sell, and everything was new. Participants learn by doing.
When a man does it, the neighbors follow. Construction is mastered, project management is familiar and sales, funds and land services are understood. Project teams move between cities. Once a project has been completed, the team may be disbanded, with people remaining in place, and people travelling to other cities to form project teams with new partners. For decades, the village community has been involved in numerous construction projects.
I asked him how you learned how to develop real estate so complicated. He said the neighbors will, slowly, too. There is a whole range of economics in this sentence. Neighbors would say that the holder of this capacity is not a company, but a village. The company was founded and dissolved, the project began and ended, and the experience in the village was not lost. What it is slowly saying is that there is both technical knowledge and role experience and common calibration methods. When a project is completed, the original specific collaborative relationship may not be fully preserved, but participants know how to repartify, validate and collaborate more quickly in the new project. What they learned was calibration procedures.
Real estate development is, of course, also dependent on land opportunities, financing relationships and project information, which are shared among the people in these villages. In order to determine which part comes from the organization's calibration, three questions will be asked in the future: how the project team is moving, what different people have learned from each other's positions, and how long a new project will take from preparation to stabilization. But the story has shown one thing: the organizational experience of a region can be built up through a network of people from the same village, family and occupation, not always by large enterprises or industrial parks. The first projects covered the cost of learning, and later people used role lessons from previous generations. This capacity is attached to a group of people, not to any enterprise.
X. Going to sea: Why do mature firms have to learn to produce again abroad?
I have been telling you this for several years: over the next two decades, there may be a large number of firms in China that can take root abroad, perhaps on a scale of over 100,000. It is not an export; it is the construction, hiring and organization of production on the land of others. This could lead to a significant movement of multinational production organizations, reminiscent of the experience of Hong Kong's Austral capital entering the Pearl Triangle in the 1980s and the presence of Japanese enterprises in North America. Why is the first factory in this paper so difficult? It's not an academic issue for those companies that are ready to go to sea, it's what they have to face tomorrow morning.
Let me ask you a strange question. A business that has been in China for 20 years has come to Vietnam or Mexico to learn how to produce again. Its equipment, drawings, process parameters and suppliers ' lists are all available. Its plant directors, workshop directors, long lines, equipment and mass skeletons can also be brought. It can even take the whole class. It does not involve local worker-side experience, local suppliers and local work environments. Productive capacity is not a list of knowledge, but the result of calibration of these and what it takes. When we get overseas, the calibration has to be repeated.
And there are more calibrations abroad than in the country. Local workers may have experience, may not, and may not be compatible with Chinese managers. Language, working-time habits, trade unions, religious festivals, legal provisions for overtime and dismissal, and the pace of delivery of local suppliers are parameters that are never calibrated in the country. Using the symbol in section VI, the K overseas is larger and the experience is less compatible, so both a and b are longer. I would call it organizational calibration distance: the compatibility of production organization between two regions. It is not geographic or cultural. Mexico and China are geographically far away, but an enterprise that has worked as an electronic agent in Dongguan is taking it to the electronics industry in northern Mexico, where calibration distance may not be significant because the worker side experience has been nurtured in factories in the United States and Asia. In turn, to a place where no modern plant has ever existed, the geographic proximity and the organization's calibration is far away. The concept is not yet a strict variable. It is a point of view that deserves to be studied: deciding whether or not tissue technology can pass the past, how quickly, may be the organization calibrating distance, not the distance on the map.
Here is an important precedent. In 1984, Toyota and General General co-organized a factory in Frimont, California. The factory was originally universal and was known for its poor quality and labour conflicts, and it was shut down. The joint venture rehired about 85 per cent of the union workers at the original plant, using Toyota production organization. By 1986, its productivity had exceeded that of other plants under the general flag and its quality had improved significantly. This example is often used to illustrate the importance of management practices. From this point of view, I'm talking about another thing: the workers are not replaced, they're re-calibrated. Toyota takes the management side of the process, the workers bring the worker side of the experience, and the two sides rematch the new parameters. The calibration takes time, but is far less than the training of workers from zero. The success of Japanese enterprises in North America in the 1980s was largely the success of this two-sided calibration.
For offshore enterprises, the next issue is the development of local managers. The logic of section VII applies here as it is. The company can take the lead and the entire team out of China, with the Chinese running the core segment and the locals on the outside. The relocation of entire groups would preserve some of the rules of collaboration among the original members and would usually help to shorten the initial wearout. However, the common calibration Q in operation in the country cannot be directly relocated abroad: the new plant will still have to be calibrated with local equipment, suppliers, personnel and the working environment. If the local population is permanently involved only in productive jobs, local experience is largely worker-side; follow-up plants may still depend on external managers. However, as long as local staff members gradually enter positions as long as they are liners, workshop directors, etc., the relocation of teams can also be the starting point for developing local management experience. Enterprises can also place local people in the positions of class leaders and section chiefs from day one, and Chinese managers can take them with them. This may increase initial training and coordination, and may also reduce certain calibrations by utilizing local familiarity with language, labour market and institutional environments. A few years later, the accumulation of local management experience will depend on what local personnel actually occupy and how long they continue to participate. What is truly worth comparing is how different staffing levels affect both current calibration speed and future organizational reproduction capacity.
If the size of the overseas establishment continues to grow, reliance on an out-posted management team may be limited by the availability of personnel, costs and local working environment. The history of the Pearl Triangle will be reversed here: At the time when the port company had made a local director, Chinese companies now had to answer the same question, except for this time standing in the position of the port company. The evolution of local management positions may affect local ownership of more similar organizations after many years. This is a matter for early reflection by offshore companies.
Two things are worth taking from this paper. One is the third construction team. A part of the enterprise ' s domestic production capacity has never been on its payroll, in suppliers, construction teams and foreign associations that it has cooperated with for many years. They're not here. Businesses find themselves smaller than they do in the country, and many of the things they didn't have to worry about are suddenly done by themselves. Rebuilding this circle of collaboration may sometimes be more time-consuming than rebuilding the workshop itself. The other is the village of section 9. The first companies to go to sea paid for the calibration of learning costs, and later they could use the people they left behind. That's why we're at the same time.
Auto-fetched and translated by AI Atlas. All rights belong to the original author.