Sensitive and Crucial: on Measuring Living Standards in the 18th Century

In the course of the twitterminar on the High-Wage Economy argument (HWE) which generated responses from John Styles on his blog (who has convinced me that the key solution to HWE rests in Normandy, not the Alsace) and many other on Twitter. In the course of that discussion, I skirted a point I have been meaning to make for a long time. However, I decided to avoid it because it is tangentially related to the HWE story. Its about how we measure living standards over space in the past.

Basically, the HWE story is a productivity story and all that matters in such a story is wage rates relative to other input prices. Because we’re talking about relatives, the importance of proper deflators is not that crucial. However, when you move beyond HWE and try to ask the question regarding absolute differences over space in living standards, the wage rates are not sufficient and proper deflators are needed.

They are many key issues to estimating living standards across space. The largest is that given that very few goods crossed borders in the past, converting American incomes into British sterling units using reported exchange rates would be rife with errors and calculating purchasing power parities would be complicated. The solution, very simple and elegant by its simplicity, is to rely on the logic of the poverty measures. Regardless of where you are, there is a poverty threshold. Then, all that is needed is to express incomes as the ratio of income to the poverty line. If the figure is three, then the average income buys three times the poverty line. Expressed as such, comparisons are easy to do. This is what Robert Allen did and it was basically a deeper and more complete approach than Fernand Braudel’s “Grain-Wages” (wage rates divided by grain prices).

Where should the line be?

While this represents a substantial improvement for economic historians like me who are deeply interested in “getting the data right”, there are flaws. In the course of my dissertation on living standards in Canada (see also my working papers here and here), I saw one such flaw in the form of how long the length of the work year was. In fact, a lot of my comments in this post were learned on the basis of Canada as an extreme outlier in terms of sensitivity. In Canada, winter is basically a huge preindustrial limitation on the ability to work year-round (thus, the expression mon pays ce n’est pas un pays, c’est l’hiver). But this flaw is only the tip of the iceberg. First of all, the winter means that the daily energy intake must substantially greater than 2,500 calories in order to maintain body mass. The mechanism through which the temperature increases the energy requirements of the human metabolism is in part the greater weight carried by the heavier clothing in addition to the energy needed by the body to maintain body temperature. At higher altitudes, these are compounded by the difference in air pressure.In their attempt to construct estimates of the living standards of Natives in the Canadian north during the fur trade era, Ann Carlos and Frank  Lewis assert that it is necessary to adjust the basket of comparison to include more calories for the natives given the climate – they assert that 3500 calories were needed rather 2500 calories for English workers.In Russia, Boris Mironov estimated that the average calories ingested stood at 2952 per day between 1865 and 1915 while the adult male had to consume 3204 calories per day. In Canada in the 18th century, it was estimated that patients at the Augustines hospital in Quebec City required somewhere 2628 calories and 3504 calories per day while soldiers consumed on average 2958 calories per day and the average population consumed 2845 calories per day (see my papers linked up above).  The range of calorie requirements for soldiers (which I took from a reference inside my little sister’s military stuff) is quite large: from 3,100 in the desert at 33 degrees Celsius to 4,900 in artic conditions (minus 34 degrees Celsius) – a 58% difference. So basically, when we create welfare ratios for someone in, say, Mexico, the calories needed in the basket should be lower than in the Canadian basket.

Another issue, of greater importance, is the role of fuel. In the welfare ratios commonly used, fuel is alloted at 2MBTU for the basic level of sustenance which. This is woefully insufficient even in moderately warm countries, let alone Canada. My estimates of fuel consumption in Canada is that the worst case hovers around 20MBTU (ten times above the assumption) if the most inefficient form of combustion (important losses) and the worst kind of wood possible (red pine). Similar levels are observed for the American colonies.

Combined together, these corrections suggest that the Canadian poverty threshold should be higher than the one observed in France, England, South Carolina or Argentina. These adjustments can more or less be easily made by using military manuals. The army measures the basic calories requirements for all types of military theaters.

How to factor in family size and use equivalence scales. 

Equivalence scales refer to the role of family size. Given the same income, families of different size will have different levels of welfare. Thanks to economies of scale in housing, cooking, lighting and heating, larger households can get more utility out of one dollar of income. That adjustments are required to render different households comparable is well accepted amongst economists. However, given the sensitivity of any analysis to the assumptions underlying any adjustments, there is an important debate to be had.

The convention among economic historians has been to assume that households have three adult equivalents. This assumption has gone largely undiscussed. The problem is “which scale to use”. The conversion into adult equivalents is subject to debates. Broadly speaking, three approaches exist. The first uses the square root of the number of individuals. The second attributes the full weight of the first adult, half the weight of the second adult and 30% for each child. This approach is commonly used by the OECD, Statistics Canada and numerous government agencies in Canada The third approach is the one used by the National Academy of Sciences in the United States which proposed to use an exponent ranging between 0.65 and 0.75 to household size but only after having multiplied the number of children by 0.7. As a result, a family of four (two parents, two infants) can have either 2 adult equivalents (square root), 2.1 adult equivalents (OECD and Statistics Canada approach) or 2.36 adult equivalent (NAS approach). The differences relative to the square roots approach are 5% and 18%. If we move to a family of 6 persons, the differences increase to 10.22% and 34.72%.  If we are comparing regions with identical family structures, this would not be a problem. If not, then it is an issue. The selection of one method over another would have important effect on the cost of the living basket, with the NAS approach showing the costliest basket. Using a method relatively close to that of the OECD (although not exactly that measure), Eric Schneider found that the relatively small size of families in England led Allen to underestimate living standards. In a more recent paper, Allen alongside Schneider and Murphy pointed out that extending Schneider’s analysis to Latin America where “family sizes were likely larger (…) than in England and British North America” would amplify the wage gap between the two regions.

familysize

The table above shows how much family size varied around the late 17th century across region. Clearly, this is a non-negligible issue.

Sensitivity of estimates

Just to see how much these points matter, let’s modify for two easily modifiable factors: household size (given the numbers above) and fuel requirements (calories from food are harder to adjust for and I am still in the process of doing that). Let’s recompute the welfare ratios (those classified as bare bones) of Canada (the outlier) relative to the other according to different changes circa the end of the 17th century. How much does it matter?

Comparing New World places like Canada and Boston does not change much – they are more or less similar (family size and relative price-wise). However, just adjusting for family size eliminates a quarter of the gap between Canada and Paris (from 61% to somewhere 43.9% and 49.5%). Then, the adjustment for the fact that it is freezing cold in Canada eliminates a little more than half the advantage Canada enjoyed. So roughly two third of the Canadian advantage over Paris (the richest place in France) is eliminated by adjusting for family size and fuel consumption without adjusting for food requirements. However, family size does not affect dramatically the comparison between Paris and London (regardless of whether we use the Allen figures or the Stephenson-Adjusted figures).  Thus, most of the sensitivity issues are related to comparing the New World with the Old World. effectofcorrections

Still, there are some appreciable differences from family structures within Europe (i.e. the Old World) that may alter the relative positions.  For example, Ireland had much larger families than England in the 18th century (see here – the authors shared their dataset with me and a co-author): in 1700, England & Wales had an average household size of 4.7 compared with 5.32 in Ireland. That would moderately disrupt the comparison. Not as much as comparison between the New World and Old World, but enough to make cautious about European differences.

Conclusion

I have seen many discussions regarding the sensitivity of welfare ratios in numerous papers. I am not attempting to make my present point into some form of revolutionary issue. However, all the sensitivity estimates were concentrated on a case or another and they all concern a specific problem. No one has gathered all the problems in one place and provided a “range of estimates”. Maybe its time to go in that direction so that we know which place was poor and which was not (relative to one another, since anything preindustrial was basically dirt-poor by our modern standards).

Advertisements

Testing the High-Wage Economy (HWE) Hypothesis

Over the last week or so, I have been heavily involved in a twitterminar (yes, I am coining that portemanteau term to designate academic discussions on twitter – proof that some good can come out of social media) between myself, Judy Stephenson , Ben Schneider , Benjamin Guilbert, Mark Koyama, Pseudoerasmus,  Anton Howes (whose main flaw is that he is from King’s College London while I am from the LSE – nothing rational here), Alan Fernihough and  Lyman Stone. The topic? How suitable is the “high-wage economy” (HWE) explanation of the British industrial revolution (BIR).

Twitter debates are hard to follow and there is a need for summaries given the format of twitter. As a result, I am attempting such a summary here which is laced with my own comments regarding my skepticism and possible resolution venues.

An honest account of HWE

First of all, it is necessary to offer a proper enunciation of HWE’s role in explaining the industrial revolution as advanced by its main proponent, Robert Allen.  This is a necessary step because there is a literature attempting to use high-wages as an efficiency wage argument. A good example is Morris Altman’s Economic Growth and the High-Wage Economy  (see here too) Altman summarizes his “key message” as the idea that “improving the material well-being of workers, even prior to immediate increases in productivity can be expected to have positive effects on productivity through its impact on economic efficiency and technological change”. He also made the same argument with my native home province of Quebec relative to Ontario during the late 19th century. This is basically a multiple equilibria story. And its not exactly what Allen advances. Allen’s argument is that wages were high in England relative to energy. This factors price ratio stimulated the development of technologies and industries that spearheaded the BIR. This is basically a context-specific argument and not a “conventional” efficiency wage approach as that of Allen. There are similarities, but they are also considerable differences. Secondly, the HWE hypothesis is basically a meta-argument about the Industrial Revolution. It would be unfair to caricature it as an “overarching” explanation. Rather, the version of HWE advanced by Robert Allen (see his book here) is one where there are many factors at play but there is one – HWE – which had the strongest effects. Moreover, while it does not explain all, it was dependent on other factors that contributed independently.  The most common view is that this is mixed with Joel Mokyr’s supply of inventions story (which is what Nick Crafts has done). In the graph below, the “realistically multi-causal” explanation is how I see HWE. In Allen’s explanation, it holds the place that cause #1 does. According to other economists, HWE holds spot #2 or spot #3 and Mokyr’s explanations holds spot #1.

hwe

In pure theoretical terms (as an axiomatic statement), the Allen model is defensible. It is a logically consistent construct. It has some questionnable assumptions, but it has no self-contradictions. Basically, any criticism of HWE must question the validity of the theory based on empirical evidence (see my argument with Graham Brownlow here) regarding the necessary conditions. This is the hallmark of Allen’s work: logical consistency. His work cannot be simply brushed aside – it is well argued and there is supportive evidence. The logical construction of his argument requires a deep discussion and any criticism that will convince must encompass many factors.

Why not France? Or How to Test HWE

As a doubter of Allen’s theory (I am willing to be convinced, hence my categorization as doubter), the best way to phrase my criticism is to ask the mirror of his question. Rather than asking “Why was the Industrial Revolution British”, I ask “Why Wasn’t it French”. This is what Allen does in his work when he asks explicitly “Why not France?” (p.203 of his book). The answer proposed is that English wages were high enough to justify the adoption of labor-saving technologies. In France, they were not. This led to differing rates of technological adoptions, an example of which is the spinning jenny.

This argument hinges on some key conditions :

  1. Wages were higher in England than in France
  2. Unit labor costs were higher in England than in France (productivity-adjusted wages) (a point made by Kelly, Mokyr and Ó Gráda)
  3. Market size factors are not sufficiently important to overshadow the effects of lower wages in France (R&D costs over market size mean a low fixed cost relative to potential market size)
  4. The work year is equal in France as in England
  5. The cost of energy in France relative to labor is higher than in England
  6. Output remained constant while hours fell – a contention at odds with the Industrious Revolution which the same as saying that marginal productivity moves inversely with working hours

If most of these empirical statements hold, then the argument of Allen holds. I am pretty convinced by the evidence advanced by Allen (and E.A. Wrigley also) regarding the low relative of energy in England. Thus, I am pretty convinced that condition #5 holds. Moreover, given the increases in transport productivity within England (here and here), the limited barriers to internal trade (here), I would not be surprised that it was relatively easy to supply energy on the British market prior to 1800 (at least relative to France).

Condition #3 is harder to assess in terms of important. Market size, in a Smithian world, is not only about population (see scale effects literature). Market size is a function of transaction costs between individuals, a large share of which are determined by institutional arrangements. France has a much larger population than England so there could have been scale effects, but France also had more barriers to internal trade that could have limited market size. I will return to this below.

Condition #1,2,4 are basically empirical statements. They are also the main points of tactical assault on Allen’s theory.  I think condition #1 is the easiest to tackle. I am currently writing a piece derived from my dissertation showing that – at least with regards to Strasbourg – wages in France presented in Allen (his 2001 article) are heavily underestimated (by somewhere between 12% and 40% using winter workers in agriculture and as much as 70% using the average for laborers in agriculture). The work of Judy Stephenson, Jane Humphries and Jacob Weisdorf has also thrown the level and trend of British wages into doubts. Bringing French wages upwards and British wages downwards could damage the Allen story. However, this would not be a sufficient theory. Industrialization was generally concentrated geographically. If labor markets in one country are not sufficiently integrated and the industrializing area (lets say the “textile” area of Lancashire or the French Manchester of Mulhouse or the Caën region in Normandy) has uniquely different wages, then Allen’s theory can hold since what matters is the local wage rate relative to energy. Pseudoerasmus has made this point but I can’t find any mention of that very plausible defense in Allen’s work.

Condition #2 is the weakest point and given Robert Fogel’s work on net nutrition in France and England, I have no problem in assuming that French workers were less productive. However, the best evidence would be to extract piece rates in textile-producing regions of France and England. This would eliminate any issue with wages and measuring national productivity differences. Piece rates would perfectly capture productivity and thus the argument could be measured in a very straightforward manner.

Condition #4 is harder to assess and more research would be needed. However, it is the most crucial piece of evidence required to settle the issue once and for all. Pre-industrial labor markets are not exactly like those of modern days. Search costs were high which works in a manner described (with reservations) by Alan Manning in his work on monopsony but with much more frictions. In such a market, workers may be willing to trade in lower wage rates for longer work years. In fact, its like a job security argument. Would you prefer 313 days of work guaranteed at 1 shilling per day or a 10% chance of working 313 days for 1.5 shillings a day (I’ve skewed the hypothetical numbers to make my point)? Now, if there are differences in the structure of labor markets in France and England during the 18th and 19th centuries, there might be differences in the extent of that trade-off in both countries. Different average discount on wages would affect production methods. If French workers were prone to sacrifice more on wages for steady employment, it may render one production method more profitable than in England. Assessing the extent of the discount of annual to daily wages on both markets would identify this issue.

The remaining condition (condition #6) is, in my opinion, dead on arrival. Allen’s model, in the case of the spinning jenny, assumed that labor hours moved in an opposite direction as marginal productivity. This is in direct opposition to the well-established industrious revolution. This point has been made convincingly by Gragnolati, Moschella and Pugliese in the Journal of Economic History. 

In terms of research strategy, getting piece rates, proper wage estimates and proper labor supplied estimates for England and France would resolve most of the issue. Condition #3 could then be assessed as a plausibility residual.  Once we know about working hours, actual productivity and real wages differences, we can test how big the difference in market size has to be to deter adoption in France. If the difference seems implausible (given the empirical limitations of measuring effective market size in the 18th century in both markets), then we can assess the presence of this condition.

My counter-argument : social networks and diffusion

For the sake of argument, let’s imagine that all of the evidence favors the skeptics, then what? It is all well and good to tear down the edifice but we are left with a gaping hole and everything starts again. It would be great to propose a new edifice as the old one is being questioned. This is where I am very much enclined towards the rarely discussed work of Leonard Dudley (Mothers of Innovation). Simply put, Dudley’s argument is that social networks allowed the diffusion of technologies within England that fostered economic growth. He has an analogy from physics which gets the point across nicely. Matter has three states : solid, gas, liquid. Solids are stable but resist to change. Gas, matter are much more random and change frequently by interacting with other gas, but any relation is ephemeral. Liquids permit change through interaction, but they are stable enough to allow interactions to persist for some time. Technological innovation is like a liquid. It can “mix” things together in a somewhat stable form.

This is where one of my argument takes life. In a small article for Economic Affairs, I argued (expanding on Dudley) that social networks allowed this mixing (I am also expanding that argument in a working paper with Adam Martin of Texas Tech University). However, I added a twist to that argument which I imported from the work of Israel Kirzner (one of the most cited books in economics, but not by cliometricians – more than 7000 citations on google scholar). Economic growth, in Kirzner’s mind,  is the result of entrepreneurs discovering errors and arbitrage possibilities. In a way, growth is a process of discovering correcting errors. An analogy to make this point is that entrepreneurs look for profits where the light is while also trying to move the light to see where it is dark. What Kirzner dubs as “alertness” is in fact nothing else than repeated and frequent interactions. The more your interact with others, the easier it becomes for ideas to have sex. Thus, what matters is how easy it is for social networks to appear and generate cheap information and interactions for members without the problem of free riders. This is where the work of Anton Howes becomes very valuable. Howes, in his PhD thesis supervised by Adam Martin who is my co-author on the aforementioned project (summary here), showed that most innovators went in frequent with one another and they inspired themselves from each other. This is alertness ignited!

If properly harnessed, the combination of the works of Howes and Dudley (and also James Dowey who was a PhD student at the LSE with me and whose work is *Trump voice* Amazing) can stand as a substitute to Allen’s HWE if invalidated.

Conclusion

If I came across as bashing on Allen in this post, then you have misread me. I admire Allen for the clarity of his reasoning and his expositions (given that I am working on a funded project to recalculate tax-based measures in the US used by Piketty to account for tax avoidance, I can appreciate the clarity in which Allen expresses himself). I also admire him for wanting to “Go big or go home” (which you can see in all his other work, especially on enclosures). My point is that I am willing to be convinced of HWE, but I find that the evidence leans towards rejecting it. But that is very limited and flawed evidence and asserting this clearly is hard (as some of the flaws can go his way). Nitpicking Allen’s HWE is a necessary step for clearly determining the cause of BIR. It is not sufficient as a logically consistent substitute must be presented to the research community. In any case, there is my long summary of the twitteminar (officially trademarked now!)

P.S. Inspired by Peter Bent’s INET research webinar on institutional responses to financial crises, I am trying to organize a similar (low-cost) venue for presenting research papers on HWE assessment. More news on this later.

England circa 1700: low-wage or high-wage

A few months ago, I discussed the work of my friend (and fellow LSE graduate) Judy Stephenson on the “high-wage economy” of England during the 18th century. The high-wage argument basically states that high wages relative to capital incite management to find new techniques of production and that, as a result, the industrial revolution could be initiated. Its a crude summary (I am not doing it justice here), but its roughly accurate.

In her work, Judy basically indicated that the “high-wage economy” observed in the data was a statistical artifact. The wage rates historians have been using are not wage rates, they’re contract rates that include an overhead for contractors who hired the works. The wage rates were below the contract rates in an amplitude sufficient to damage the high-wage narrative.

A few days ago, Jane Humphries (who has been a great inspiration for Judy and whose work I have been discretely following for years) and Jacob Weisdorf came out with a new working paper on the issue that have reinforced my skepticism of the wages regarding England. A crude summary of Humphries and Weisdorf’s paper goes as such: preindustrial labor markets had search costs, workers were willing to sacrifice on the daily wage rate (lower w) in order to obtain steady employment (greater L) and thus the proper variable of interest is the wage paid on annual contracts.

While their results do not affect England’s relative position (it only affects the trend of living standards in England), it shows that there are flaws in the data. These flaws should give us pause before proposing a strong theory like the “high-wage economy” argument. Taken together, the work of Stephenson (whom I am told is officially forthcoming), Humphries and Weisdorf show the importance of doing data work as the new data may overturn some key pieces of research (maybe, I am not sure, there is some stuff worth testing).

The High Wage Economy: the Stephenson critic

A recent trend has emerged in economics. The claim is that high wages can have a dynamic positive effect on market economies.  The intuition is that high wages increase productivity because they incite management to find new techniques of production. In essence, its an argument about efficiency wages: efficiency wages increase incentives to innovate on the part of managers, they can also incite workers to acquire more human capital and work harder and more diligently.

In economic history, this claim has been taken up by scholars like Robert Allen (see his work here for the general public) who argues that the Industrial Revolution took place in England because of high wages. The high-wages of England in the 17th and 18th centuries (relative to all other areas in Europe), together with cheap energy, created an incentive for capital-intensive methods of production (i.e. the industrial revolution). In fact, a great share of the literature on the desirability of high wages for economic development has emanated from the field of economic history.

I have always been skeptical of this argument for two reasons. The first is that efficiency wages is a strange theory that relies on debatable assumptions about labor (strangely, I have been convinced of this point by Austrian scholars like Don Bellante and Pavel Ryksa). The second is that numerous scholars have advanced large criticisms of the underlying data. Robert Allen – the figurehead proponent of the high wage argument – has been constantly criticized by historians like Jane Humphries (see here) for the quality of the data and assumptions used. Allen defends himself on numerous occasions and many of his replies (mainly those on the role of family size in living standards) show that his initial case might have been too conservative (i.e. he is more “correct” than he claims).

Until a year or two ago, I was agnostic on the issue even though I was skeptical. That was until I met Judy Stephenson – a colleague at the London School of Economics. Judy did what I really like to do – dig for data (yes, I am weird like that). She went to the original sources of data used by Allen and others and she looked at what any Law-and-Economics buffs like me like to look at – transaction costs and contracting models.

She recently published her work as a working paper at the LSE and what she found is crucial! Labor was not hired directly, it was hired through contractors who charged costs on the basis of days worked. But this did not translate into wages actually paid to workers. The costs included risks and overheads for contractors. Somewhere between 20% and 30% of the daily costs were not given to workers as wages. Thus, the wage series used to claim that England (Stephenson concentrates on London though) had high wages are actually 20% to 30% below the level often reported. They are also substantially close to those in western Europe.

Thus, the high wage story for England seems weaker. This little piece of historical evidence brought about by Judy is something to think about carefully when one makes the argument that high wages are conducive to growth. Since most of the argument brought to the public was informed largely by this argument in economic history, it makes sense to be cautious when thinking about it in the future.