經濟學中的最大未知數 The greatest unknown the impact of technology on the economy

Economists have always recognised that the long-run Growth of productivity is, in the end, almost the only thing that matters for the living standards of the population as a whole. Recently, there have been significant downgrades to consensus estimates of productivity growth which, if maintained for long, would have enormous effects on the attainable level of gross domestic product per capita.

經濟學家一貫認爲，從根本上來說，生產率的長期增長几乎是唯一影響全體人類生活水平的因素。近來人們對生產率增長的普遍預期大幅下降，這一趨勢如果長期保持下去，將對人均國內生產總值(GDP)的可實現水平造成巨大影響。

But the future impact of technology on long-run growth is one of the great unknowns — perhaps even the greatest — in economics.

但未來技術對長期增長的影響是經濟學上的一大未知數——有可能還是最大的未知數。

An excellent example of this uncertainty occurred at the FT Business of Luxury Summit last week, in contributions from Johann Rupert and Martin Wolf. The former painted a picture of unprecedented technical advance, quoting examples from The Second Machine Age by Erik Brynjolfsson and Andrew McAfee. This book has captured the imagination by describing a future in which machine learning increases at exponential speed, rapidly replacing human skills in large parts of the economy. In a world of robot technology, driverless cars and delivery-by-drones, measured productivity growth would surely advance very quickly.

在近期英國《金融時報》奢侈品行業峯會(FT Business of Luxury Summit)上，來自約翰•魯珀特(Johann Rupert)和馬丁•沃爾夫(Martin Wolf)的觀點就充分體現出了這種不確定性。魯珀特援引安德魯•麥卡菲(Andrew McAfee)和埃裏克•布林約爾鬆(Erik Brynjolfsson)合著的《第二次機器革命》(The Second Machine Age)中所舉的例子，描繪了前所未有的科技進步的景象。這本書描述的未來激起了人們的無限遐想：機器的學習能力以指數速度增長，迅速在經濟的大部分領域取代人類技能。在一個機器人科技、無人駕駛汽車和無人機送貨的世界裏，人們衡量的生產率增長必定前進得非常快。

Martin Wolf, however, disagreed. He argued that the great technological advances of the 19th and 20th centuries would not be replicated in the future, so productivity growth would remain subdued, as it has been since about 2003. Sympathising with the somewhat gloomy paper published by Robert Gordon in 2012, Martin felt that the low-hanging technological fruits had already been picked, and that the period of rapid advance that ended in the early 1970s was an aberration.

馬丁•沃爾夫不認同這一點。他認爲19世紀和20世紀那種技術大進步無法在未來複制，因此生產率增長將像大概2003年之後一樣，持續處於低迷狀態。他與羅伯特•戈登(Robert Gordon)2012年發表的一份有些悲觀的報告意見一致，馬丁感到，容易摘得的技術果實已經被摘下，上世紀70年代初結束的那段技術飛速進步的時期具有偶發性。

Who is right?

誰對誰錯？

My own response to Martin was to suggest that he might be under-estimating the seismic change that is taking place in the availability and communication of all types of information, as represented for example in the arrival of the iPhone. This single product has increased consumer well-being in ways that are difficult to capture in official economic statistics. Many of the key consumer advances that have taken place in the past few decades can now be done better, or more conveniently, by a single gadget. Almost none of this is picked up by the consumer price index, which continues to measure all these items separately, while not giving any weight to the extra convenience now freely available to the consumer.

對於馬丁，我認爲他可能低估了各種類型的信息的獲取和交流正在發生的重大改變，iPhone的問世就體現出這一改變。這款產品提高消費者福祉的方式難以用官方經濟數據反映。過去幾十年中在消費者領域發生了許多關鍵性進步，現在只靠一個設備就可以達到更好的效果，或者更方便地達到目的。消費者價格指數(CPI)幾乎不會計入這一切，CPI一直單獨衡量所有的項目，不會將現在消費者免費享用的額外便利納入考慮。

What is more, most individuals in advanced economies now have greater and easier access to information than any individual, however elevated, had in (say) 1990. Soon, that advantage will also apply to most citizens in emerging economies, which will therefore leap-frog the legacy technologies (for example. telephone wires) of the advanced world.

此外，現在發達經濟體的大多數個人比過去的任何人都能更便利地接觸到更多信息，哪怕過去（比如1990年）的這個人多麼位高權重。很快，新興經濟體的大多數公民也將擁有這一優勢，進而超越發達國家的傳統技術（比如電話線路）。

In his 2008 paper Ideas and Growth, Robert Lucas persuasively argued that productivity growth in advanced economies in recent centuries “is mainly an intellectual achievement, a sustained flow of new ideas” (see also the videohere). Any way of improving the flow of ideas increases the pace at which individuals can learn from others who are smarter or better informed than themselves. In these models, the propagation of knowledge is a prime determinant of growth, and that propagation is surely changing at a dramatic pace at present.

羅伯特•盧卡斯(Robert Lucas)在其2008年的報告《思想和增長》(Ideas and Growth)中進行了頗具說服力的論述。他認爲近幾個世紀以來，發達經濟體的生產率增長“基本是智力的成果，來自於新思想的持續流動”。思想流動的任何改善會都加快個人從其他更聰明或更博識的人那裏學習的步伐。在這些模型中，知識的傳播是推動增長的最重要決定因素，而這種傳播現在無疑正以極快的速度發生改變。

It would be great if we could accurately measure the impact of technological change on actual productivity growth. Some of this is relatively easy. For example, it is nowfairly clear that the temporary surge in productivity from about 1995-2003 depended largely on a surge in IT investment by a group of mainly service industries. But more often, direct measurement is extremely difficult.

如果我們能夠精確衡量技術改革對實際生產率增長的影響就好了。有時衡量這種影響相對簡單。比如，1995年到2003年間生產率的短暫大幅上升，主要依靠以服務業爲主的一些行業對IT投資的激增。但一般來說，直接衡量這種影響極爲困難。

In a 1998 paper, which deserves much wider recognition in the public debate, William Nordhaus showed how mismeasurement of the price of light, or illumination, has caused important distortions in official economic data for inflation and productivity since 1800. Mr Nordhaus simulates how the “price of light” might have been estimated in official consumer price index statistics, had they existed at the time.

經濟學家威廉•諾德豪斯(William Nordhaus)1998年的一份報告理應在公開辯論中得到更廣泛的承認。他在報告中證明了對光，或者說照明的價格的錯估，使1800年以來官方對通脹和生產率的經濟數據產生了重大扭曲。諾德豪斯模擬瞭如果當時存在官方CPI數據統計，在CPI中“光的價格”可能得到怎樣的估算。

He proxies this by measuring the price of gas and electricity (“official” price estimates I and II in the graph), which would have been the inputs used as proxies for the price of light in standard CPI indices. He then compares this with the “true” price of light, which allows for the vast improvement in the efficiency of lighting technology over time. The difference is enormous: the “official” CPI would have over-estimated the inflation rate in the provision of light by 3.6 per cent per annum for many decades. Since light accounts for about 1 per cent of consumer spending, this alone means that the change in real GDP per capita was underestimated by about 7 per cent over two centuries.

他的模擬方式是衡量天然氣和電力（右圖中的“官方”價格I和II）的價格，在標準的CPI指數中，把這兩者當做光的價格的替代品輸入其中。然後他將這與光的“真實”價格進行比較，後者考慮了隨時間推移照明技術能效的巨大提高。差異是巨大的：數十年裏，“官方”CPI每年對照明方面的通脹率會高估3.6%。由於照明在消費者支出中約佔1%，僅這一項就意味着實際人均GDP在兩個世紀以來被低估了7%左右。

If we make an analogy between the price of light in the last two centuries, and the price of information in this century, the impact of IT and the smartphone on inflation may be severely mismeasured, in which case real income and productivity may already be growing much faster than officially estimated (see Martin Feldsteinhere).

如果我們把過去兩個世紀的光的價格和本世紀信息的價格做一個類比，我們可能嚴重錯估了IT技術和智能手機對通脹的影響——實際收入和生產率或許比官方估計的增長得更快。

Unfortunately, this single example does not prove that, taken overall, technology will contribute to faster increases in productivity in the period ahead. Mr Nordhaus uses his estimates on the price of light to make a very rough guess of the total distortion to the overall CPI that stems from similar types of mismeasurement, in which improvements in quality, and the arrival of new products, are omitted from the inflation data.

遺憾的是，僅這一個例子還無法證明，整體而言，技術將使生產率在接下來的時期中增長更快。諾德豪斯通過對光的價格的估計，對整體CPI的總體扭曲程度進行了非常粗略的推測，這些扭曲來自於類似的錯估——通脹數據遺漏了質量的提升和新產品的問世。

He says that two-fifths of the calculations within the CPI are “virtually useless”, and his guess is that real incomes and productivity have been growing at 0.5-1.4 per cent per annum faster than shown in the official data — a huge potential error.

諾德豪斯表示，CPI中有五分之二的計算“幾乎無用”。他推測，每年實際收入和生產率的增長比官方數據所顯示的快了0.5%-1.4%，這是一個潛在的巨大錯誤。

Conclusion

結論

In order to be either optimistic or pessimistic about the great technology question, we therefore need to take a view on the future pace of all technical changes, compared with the sum total of all past changes. It is not enough to point to one or two dramatic new advances, however spectacular they may be.

爲了對技術這個重大問題形成一種態度，無論是樂觀的還是悲觀的，我們需要考慮未來所有技術變革的步伐，並將其與過去一切變革的總量相比較。只針對一兩次巨大的進步是不夠的，不管它們多麼引人注目。

So, while I remain optimistic that the collapse in the price of information will greatly advance productivity growth, it is hard to be completely sure that this will outweigh all other factors.

因此，儘管我對信息價格的大幅下滑將極大地促進生產率增長保持樂觀，但很難完全確定其作用是否會超過所有其他因素。

Only the central banks and the government statisticians have the scale of resources needed to solve this empirical conundrum, and currently they seem reluctant to address the issue. Traditional statistical methods, however badly flawed, rule the roost.

只有央行和政府統計學家擁有解決這個實證難題所需的大量資源，而目前他們似乎不太情願解決這個問題。傳統的統計方法不管缺陷多麼嚴重，依然將佔據主導地位。