Terminating NAFTA: What would the impact on Canada be?

  • As NAFTA negotiations become increasingly fraught, the risk of a complete termination by US President Donald Trump is increasing
  • The long-run impact of such a termination is likely to be limited, based on the MFN tariffs the US could apply
  • The most likely impact would be in terms of the real economy, as the shock generated would hit consumption, spook investment and delay the Bank of Canada’s hiking cycle.
  • To investigate the magnitude of this effect, we model the impact on real economic variables and interpret the impact on the Canadian dollar. Our results suggest that a 6% depreciation of the Canadian dollar against the US dollar could be justified.

The election of Donald Trump to the presidency of the United States was always likely to place the North American Free Trade Association (NAFTA) under some strain. Notwithstanding the near consensus amongst the economics profession about the benefits of free trade (at least amongst developed economies), or the evidence that the decline in manufacturing jobs began well before the advent of North American Free Trade, Trump repeatedly railed against the trade deal as a cause of the unemployment and poverty which afflicts much of the US, including many of his supporters.

At the heart of the US’ dispute with NAFTA is differing conceptions of the benefits of international trade. Trump and his supporters take an increasingly mercantilist approach to trade, viewing current account deficits as a failure (rather than the outcome of a set of mutually beneficial transactions) and (increasingly unsustainably in a world where the US’ global power is on the wane) desiring powerful national dominance over the terms on which they trade. This is at odds with the trend towards supranational institutions which encourage the multilateral removal of trade barriers, and the pooling of sovereignty in common dispute resolution mechanisms. This is reflected in the US’ demands in the NAFTA negotiations so far, which have included revisions to rules of origin, and to NAFTA’s dispute resolution mechanism. These are designed to block cheap imports, and regain US Department of Commerce control over tariff rates, respectively. It is not surprising, given this, that discussions have made little progress and US negotiators are said to be becoming increasingly frustrated.

This is the context in which Canadian civil servants are said to be increasingly concerned that Trump will announce the US’ withdrawal from NAFTA. This would not take immediate effect, since he would have to give six months’ notice, and in US law it is ambiguous whether it would require congressional approval to withdraw. However, if Mexico follows through with its threat to walk away from the negotiations in this case, the impact would as good as kill the trade deal over the long run.

So what would be the impact? This depends, on whether the US reverts to the pre-existing US-Canada Free Trade Agreement or rescinds this as well. In a worst case scenario, where the US simply reverts to Most-Favoured Nation (MFN) tariffs, Canada would be much more badly hit: approaching 70% of its trade is with the US and Mexico, with the bulk of this being with the US. However, the magnitude of the pure tariff effect would likely be limited: the US’ average tariff under this scenario would be around 3.5%, and Canada’s around 4.1%, unlikely to lead to significant reductions in the volume of trade between the economies. Furthermore, by sector, few of Canada’s most significant exports would attract a large MFN tariff. The main source of long-run uncertainty is the extent to which the US, free from the shackles of NAFTA’s chapter 19, would then use countervailing tariffs (supposedly aimed at neutralising subsidies) to hit back at Canadian exporters they perceive to be unfairly supported by the government. Two relevant considerations here are, firstly, that little use of chapter 19 has been made by Canada in the last ten years, and secondly, that Canada would still have recourse to the dispute mechanisms of the WTO (though these do not include the possibility for compensation in the case of a successful case). However, a recent string of ad hoc duties and restrictions levied by the US Department of Commerce do suggest the potential for a more combative trade policy.

The short-run impact would likely be more significant with the Canadian economy potentially taking a significant hit. Consumption would suffer the most, since the impact of MFN tariffs on Canadian consumers could be of the order of 2% of current consumption (contrary to popular belief, the first people to suffer when an economy introduces tariffs are its own consumers). Naturally, Canadian exporters would face reduced US demand from the tariffs, and consequently, business investment would be muted.

To assess the magnitude of the impact on the currency, we employ a Behavioural Equilibrium Exchange Rate (BEER) model which describes the relationship between exchange rates and fundamental variables. Specifically, we assume that the CADUSD exchange rate is determined by relative productivity differentials, real rate differentials, and terms of trade differentials.

In order to model the impact of a NAFTA-based shock on the Canadian economy as a whole, we use a combination of calibration and estimation of a set of equations describing the relationships between fundamental variables. Specifically, we estimate inflation using an expectations-augmented Phillips curve (where inflation is influenced by the rate of economic growth, and expectations about future inflation); we estimate consumption and investment functions which depend on the stance of monetary policy and expectations about future economic growth, and estimate the path of net exports based on terms of trade and the movement of the real exchange rate. The stance of monetary policy is assumed to follow a Taylor rule, based on inflation and growth. Implicitly, the models assume productivity growth (and therefore the growth of potential output) and government spending are unaffected over the period we analyse, and that the Bank of Canada will always set interest rates to bring inflation to target and output to its potential level over time.

In order to simulate the effect of termination of NAFTA, we assume shocks to consumption, investment and exports commensurate with the magnitude. In scenario 1, Trump announces the withdrawal from NAFTA by March 2018 and finally withdraws in December 2018, causing a hit to consumption equal to the new tariff burden which is imposed on consumers. Exports are hit by an amount determined by the trade volume elasticity to a change in price, again in December. In March 2018, anticipating uncertainty ahead, investment slows by 5%. In scenario 2, we test the impact of a more severe outcome, with investors making additional reductions to the tune of another 5%, and consumers additionally reducing their spending by 0.5% in December.

The results are in figure 1. In our “base case” i.e. how the model interprets the current state of the economy, growth moderates slightly from its current rate, and the Bank of Canada begins to hike rates over 2018 (this is largely priced in rates futures, so the effect on 10 year rates is limited). This causes inflation to settle around its 2% target. The result is a moderate strengthening of the Canadian dollar.

Compared to this, the two scenarios we consider see growth beginning to tail off from March 2018, and turning negative in December as the NAFTA shock hits. The impact is a tailing off of inflation as spare capacity in the economy expands. The Bank of Canada responds with a lower path for real rates. The result is that, by December 2018, Canada will experience between a 3 and 6% depreciation of the Canadian dollar against the US dollar. Note that the model predicts a rapid recovery of GDP as a result of the interest rate cut, and GDP growth will then overshoot as the Bank looks to recover lost output.

Figure 1 Source: Record

Some aspects of this model are highly stylised. For example, there are various constraints which might prevent the Bank of Canada from lowering rates as in our scenarios, and our simple rules-based policy is a simplification purely for the purposes of modelling. In particular, the Bank of Canada has no history of lowering the interest rate below zero, which would present a psychological barrier for policy makers. The paths of inflation and GDP are influenced by myriad factors not considered in this model. Perhaps the least predictable aspect of the model is the exchange rate impact. Though a BEER does a reasonable job of explaining currency movements most of the time, there are occasions of significant deviation historically. In particular, if FX markets are spooked by the NAFTA withdrawal, the currency could depreciate by more than implied by fundamental factors.

What this does show us, however, is the broad magnitude of the effects of a shock NAFTA termination on the Canadian economy. The short-run impact could be relatively pronounced, and certainly has the potential so slow growth, or even in an extreme scenario, produce negative growth. A depreciation of the Canadian dollar in the region of 3-6% can be expected, but market sentiment could easily make for a more volatile, extreme reaction.

The divisible labour market? Thoughts on Haldane’s speech on Labour Markets

  • The Bank of England’s chief economist Andy Haldane suggested recently that changes in labour market structure in the UK have led to a flattening of the Phillips Curve
  • In this blog, we discuss the plausibility of this argument applied to a global context over a longer time horizon

The Bank of England’s Chief Economist Andy Haldane made headlines recently for a speech delivered in Bradford. Markets reacted to an unexpectedly hawkish speech from the MPC’s most prolific dove, who cited improvements to global growth and inflation expectations in support of moderate tightening.

However, Haldane also had some interesting comments to make about the UK labour market. He addressed the question of the UK Phillips curve (variously, the relationship between unemployment and wage growth, the relationship between unemployment and inflation, and the relationship between output and inflation). Why has wage growth been so weak, even as unemployment has fallen to record lows? Haldane’s suggestion is that a series of labour market trends have caused workers to become more “divisible”. An increase in competition amongst workers makes it difficult for them to raise wages even in very tight labour markets. The trends he cites are:

  • Decreasing unionisation
  • Rising self-employment
  • Rising numbers of zero-hour contracts
  • Rising numbers of temporary and part-time jobs

The argument is interesting, and has some important global applications. The global Phillips curve has been flattening since the 1960s. This poses difficult dilemmas for central banks who are being sent different signals from inflation and unemployment numbers. If the relationship were to weaken too much, the rationale for inflation targeting would be called into question.

Some of Haldane’s argument is highly specific to the UK. There has not in general been increase in self-employment elsewhere in the developed world (a fact which should provoke some reflection on the importance of the so-called “gig” economy). The focus on zero-hour contracts is also likely a UK-specific concern. However, there has been a significant rise in part-time, temporary and insecure work, all of which is well-known and documented. The entry of women into the workforce was likely significant in these trends, as they are more likely to have part time or temporary jobs.

One measure which captures the shift from a highly centralised labour market to a more competitive “divisible” economy is the Penn World Table’s “average annual hours worked by persons engaged” statistic (AAH), which has the advantage that it offers a long time series dating back to before the decline in the slope of the Phillips curve. This measure should fall as markets become more “divisible” as more temporary and part-time work becomes ubiquitous.

The AAH DM average is plotted against the DM average Phillips curve slope in Figure 1. Two things are immediately obvious:

  • The development of labour market divisibility has occurred over a long period of time across the developed world
  • The trend in average working hours roughly matches the trend in the Phillips curve relationship with regard to its direction over time and the period over which it has happened

None of this is to point directly to a causal relationship. However, the argument of Haldane’s speech looks plausible. The factors he cites are present across the developed world, and developments in these factors coincide with developments in the Phillips curve relationship. This is certainly an avenue worth exploring.

Source: Penn World Table, Bloomberg, Record

Why the U.S. labour market is not as tight as it looks


From a tightening labour market to inflation acceleration – how far left to go for the U.S. economy? (part 1 of 2)*

Since the Global Financial Crisis (GFC), US labour force participation has declined materially, suggesting a sizeable amount of additionally available labour supply, that is not fully captured in unemployment statistics, and considerable slack in the labour market. Indeed, President Trump has suggested that up to 100 million people not in employment could be put back to work.

35% of the participation decline can be attributed to the ageing demographics effect (“net ageing”), as the baby boomer generation reaches retirement age; a powerful secular trend that will continue to weigh on employment rates over the coming years. But, this blog finds that the shadow labour force (discouraged workers) also accounts for a large part of the shift in employment dynamics, and represents a significant source of transitory slack. If the shadow labour force reverted back to a level equivalent to 2007, participation could be boosted by 50 bps, with 800,000 people added to the workforce. Indeed, if the proportion of unemployed and discouraged workers were both to revert back to 2000 levels, around 2.6 million people could return to work, but this could take years to achieve. While it is plausible that a small potential supply of labour could be additionally found in groups of people currently in education, with family responsibilities or classified as disabled/ill, it is relatively unlikely.

100 million people returning to work is clearly implausible, but considerable slack does remain in the US labour market. Without knowing the true natural rate of unemployment, we cannot know for certain how much of the current ‘underemployment’ rate represents slack, but returning even 3 million Americans back to employment in a swift manner will represent a significant challenge for the Trump administration. Considerable labour market slack remains, and it could take several years to close the gap.

(*Part 2 will address the sensitivity of wages and prices to labour-market slack, and what this means for U.S. inflation and monetary policy.)

The lacklustre US labour-market recovery

Since the GFC, US employment dynamics have shifted considerably. The labour force participation rate has fallen by more than 3%, and has only curbed its decline in the past couple of years (Figure 1). Meanwhile, employment-to-population ratios have recovered only a small part of their 5% declines from the crisis, and still remain a long way from their pre-GFC levels (Figure 2).

Figure 1: Participation rate has collapsed…                                  Figure 2: …as has employment-to-population

Sources: Bloomberg, Worldbank

Despite the official unemployment rate having returned to pre-crisis levels (4.3% in May 2017), the above two measures suggest that there may still be considerable slack left in the US labour market. This blog sets out to explore the extent to which this may be the case.

An ageing population to blame?

Many have been quick to blame the slow labour-market recovery on demographics. One study by the Chicago Fed[1] suggested that around half of the post-1999 decline in the participation rate could be explained by long-running demographic patterns (and they are expected to continue). But is this an oversimplification of the problem?

There are some crude methods to strip out these demographic effects. One could look at the participation rate or employment-to-population ratio, for the 25-54 demographic only (Figure 3). Both data series show that, even when accounting for demographic shifts, (a) there is a significant gap yet to be bridged back to pre-crisis levels and (b) the trend is currently in that direction (back to pre-crisis). Indeed, to Neel Kashkari – President of the Minneapolis Fed and sole dissenter in the FOMC’s March vote – these measures are a clear sign of labour-market slack, and form a key part of his justification for voting against a further Federal Funds rate hike[2].

So, while a large part of the decline in participation is indeed likely due to ageing, other important factors are not always given full due consideration (notably, the effect of discouraged workers leaving the labour force, who could yet re-enter as economic activity picks back up).

Figure 3: A large employment gap yet to be closed

Sources: Macrobond, Bureau of Labor Statistics Current Population Survey

100 million Americans waiting on the sidelines?

In the run up to the US election, Donald Trump repeatedly claimed that there were 100 million people not in employment in the US. It was time to get Americans “off of welfare and back to work”! So who are these 100 million?

The oft-mentioned number refers to the sum of those officially “unemployed” (7 million) plus those ‘not in the labour force’ (95 million). In other words, it is the entire proportion of the US civilian non-institutional population (15+) that is not currently working, equal to around 40% (Figures 4 and 5).

Figure 4: 100m not in employment…                                           Figure 5: …40% of population (15+)

Sources: Bureau of Labor Statistics Current Population Survey.

The number is expectedly large, and has been growing for much of the past decade, but it includes a whole range of non-employed adults; from students to the retired, to those that are disabled or suffering from long-term illness. So it is unrealistic to expect all of these people to get a job. Instead, a detailed dissection of the numbers is essential.

Breaking it down

Figures 6 and 7 show a breakdown of those 95 million not in the labour force, while Figures 8 and 9 show the contribution from each segment to the change in the participation rate between 2007 and 2016. In total, there are five categories that could be driving spare capacity: (1) ageing/retirement, (2) in school or training, (3) disabled or ill, (4) ‘want a job’ (the shadow labour force), and (5) family responsibilities. So let’s take each one in turn…

Figure 6: Breakdown of ‘not in labour force’                                  Figure 7: Breakdown by age group

Sources: Federal Reserve Bank of Atlanta. As at December 2016.

Figure 8: Nine-year participation rate contribution                          Figure 9: Nine-year participation rate contribution

Sources: Record Currency Management, Federal Reserve Bank of Atlanta

  1. Ageing:

Retirement alone accounts for 27 million of those ‘not in the labour force’, and “ageing” represents some 210 bps of the participation rate decline over the past nine years – or 63% of the total change – perhaps unsurprisingly, as the baby boomers approach retirement age. But this number can be misleading; during the same period, it was offset by an increasing propensity to delay retirement until later in life, which actually boosted labour force participation by some 90 bps (Figure 8). Consequently, the net ageing effect in the last decade is closer to 120 bps (Figure 9). This is still the single largest contributor to the participation decline, but only accounts for 35% of it.

The ageing demographics effect is highly unlikely to reverse. 57% of all retirees are already 66+ (Figure 10), while 90% of the net-retirement effect on participation (2007-2016) has been driven by people in this age group. A counterbalancing, “delayed-retirement” effect is indeed at play, with a huge number of Americans now working much longer than past generations did. But this is already accounted for in the above net numbers. And, moreover, the vast majority of the delayed-retirement effect has so far come from the 56-65 demographic, with relatively few from the 66+ bucket delaying enough to offset demographic shifts (Figure 11).

Figure 10: Age distribution of retirees                                           Figure 11: Delayed retirement cannot offset ageing

Sources: Record Currency Management, Federal Reserve Bank of Atlanta

All in all, the data indeed suggest that delayed-retirement trends make future ageing effects less worrying than they would otherwise be. But it would be unrealistic to expect an acceleration in delayed retirement large enough to completely offset the ageing process and drive the participation rate higher. The majority of those who have already retired are not likely coming back, and ageing trends are set to continue for some time (Figure 12).

Figure 12: Ageing trends set to continue for decades

Sources: Record Currency Management, Federal Reserve Bank of Atlanta, U.S. Census Bureau. (Projected percentage based on U.S. Census Bureau rate of increase from 2015.)

  1. In school or training:

An increase in the rate of those in school or training accounts for some 90 bps of the participation decline. There are 13 million now in this category, with the vast majority aged 16-25 (Figure 13); and, unsurprisingly, the majority of the effect on participation also comes from this age bucket (Figure 14).

Figure 13: 16-25s account for most in education…                        Figure 14: …and majority of rise in education rate

Sources: Record Currency Management, Federal Reserve Bank of Atlanta

So could the increase in the education rate be a relic of the GFC, soon to revert as more jobs become available? Also unlikely.

Figure 15 shows that the percentage of 16-to-24-year-olds in school or training has been in a rising trend for quite some time. So the recent increase, since 2008, is not necessarily GFC related; in fact the rate of growth has been slower than pre-crisis. The 25+ age group, on the other hand, did see a disproportionate education-rate rise after 2008; but the absolute effect was small (10-20 bps at worst), and much of this spike has now been unwound, to revert close to the longer-term trend.

Figure 15: A long-term rising trend in the education rate

Sources: Federal Reserve Bank of Atlanta

Ultimately, all age groups have seen a long-term upward trend in school and training, suggesting that much of this change is secular in nature and not likely to revert. It is fair to assume that there will always be a certain proportion of people in education, and this proportion can be expected to rise as technological automation shifts the demand for labour toward higher-skill-based jobs. With this in mind, it is unlikely that the Trump administration can shift a significant proportion of people from education back into work.

  1. Disabled or ill:

People registered as disabled or ill account for 20 million of those not in the labour force and for 65 bps of the 9-year participation decline. While the distribution across age groups is fairly wide (Figure 16), the recent effect from an increased disabled-or-ill rate is largely driven by those aged 51 and over (Figure 17).

Figure 16: A wide distribution of disabled or ill…                           Figure 17: …but recent effect driven by 51+ group

Sources: Record Currency Management, Federal Reserve Bank of Atlanta

In fact, disability rates have been on the rise for some time, for all ages up to retirement age (Figure 18). Some might argue that this could be driven by an increase in fraudulent benefit claims, at a time when it is difficult to find a job. But any increase in fraudulent claims (if true) would be unlikely to amount to a large part of the disability rate effect; and the observed upward trend has largely continued in the past few years, even as the jobs recovery has been underway.

While it may be tempting to suggest that many classified as disabled or ill could be put back to work, 68% of the rise in this group since 2007 has come from the 55-64 age bucket. Whatever injuries they may or may not have, it is surely unlikely that a large proportion of them will come back to work before retiring.

Figure 18: Disability rates have been on the rise for some time

Sources: Federal Reserve Bank of Atlanta

If the rising trend in disability rates could be curbed, and the next wave of 55-64 year olds could revert to a disability rate two percentage points lower (the 2007 level), some 50 bps could theoretically be added back to the participation rate over the course of the next decade. But either way, in the short-to-medium term, the US administration would do well not to rely on getting this group “back to work”.

  1. Want a job (shadow labour force):

Beyond the 7.5 million officially recognised as “unemployed”, there are another 8 million in the “shadow labour force”, an important measure of underemployment. These are people who want a job but are not included in the official labour force. And this group strongly contributed towards the decline in labour force participation, particularly up to 2012 (a 100 bps contribution).

To be considered unemployed by the BLS, a person has to not only want a job but also be (1) available to work and (2) actively seeking employment. Some do not meet these criteria but are still relevant to labour-market slack statistics. For example, if an individual has actively sought work sometime in the previous year but not in the past four weeks, she is considered a “discouraged worker” who is only “marginally attached” to the labour market. She is part of the ‘shadow labour force’.

Various studies have found that the difference between the officially unemployed and others that want a job is not especially relevant; “all tend to search for work or find jobs at a higher rate than others outside of the labor force” (Atlanta Fed, 2017[3]). For that reason, this particular segment is by far the most important. All members of the shadow labour force essentially represent spare capacity in the US economy and, with the right allocation of resources, could theoretically be put to work in times of booming economic activity.

The shadow labour force is dispersed across all age groups, but especially weighted towards younger generations (Figure 19), with the 9-year participation rate effect also spread relatively evenly across all ages. Unlike the previous three non-participation categories, though, the shadow labour force has actually been in decline since 2012 (Figure 9). That is, while many fell out of the labour force into this category during 2007-2012, a large proportion of them have now come back to the workforce.

Figure 20 shows the effect of the shadow labour force on participation from 2007 to (a) 2012 and (b) 2016. In most age groups, half of the effect has already been reversed, as the job recovery has picked up momentum. And, as one might expect, the younger generations appear to be the most likely to be brought back in to the labour force (with older generations perhaps more likely to suffer from the adverse effects of not being in employment for a prolonged period of time).

Figure 19: A young shadow labour force…                                   Figure 20: …that is getting back to work

Sources: Record Currency Management, Federal Reserve Bank of Atlanta

Most encouraging of all, the rate of 16-20-year-olds in the shadow labour force has reversed entirely. At best, this may suggest that the jobs recovery could rapidly filter through, up the age groups, as the demand for labour continues to rise – potentially boosting participation by 50 bps or more. At worst, the younger generation are no longer likely to suffer as much as their predecessors from the acute lack of skills that has plagued those sitting on the side-lines for a prolonged period. And this means that higher rates of employment can now more easily persist, as these youngsters rise through the age groups in the long-run.

  1. Family responsibilities:

Finally, while some have also commentated on the likely contribution of stay-at-home parents to declining participation, this is the factor that least stands up to scrutiny. As Figures 8 and 9 show, reduced family responsibilities have actually contributed to a slight boost to the participation rate over the past nine years, with female participation benefiting the most.

So, this has not been a contributing factor over the past decade. But could it be a source of further labour market supply in the future? While one can safely assume that it will always be necessary for a certain proportion of the population to remain in this segment, there is some merit to the argument that the proportion could, hypothetically, be lower. With 25 million people currently staying at home for family responsibilities, even a 5% decline in this group – spurred by cheaper childcare measures, for example – could lead to more than 1 million joining the workforce. Ultimately though, the prospects of such a huge shift in behaviour are hard to measure and, in order to occur in the next few years, would likely require vast government-subsidy initiatives (of the likes there is no evidence to suggest will transpire) combined with a significant cultural shift on the part of the American people.

The ‘real’ rate of unemployment…

In summary, we have learned that much of the decline in the labour force participation rate since 2007 is unlikely to be fully reversed in the short-to-medium run – there are powerful secular trends at play.

But, equally, some of the decline is transitory. Beyond simply focussing on the official unemployment rate, discouraged workers in particular represent a significant addition to labour market slack. Combining 7 million unemployed with 6.5 million discouraged workers means that around 13.5 million people currently want a job (the ‘real’ rate of unemployment). And, if the shadow labour force was to revert back to a level equivalent to 2007, participation could be boosted by 50 bps, with 800,000 people added to the workforce.

A broader BLS measure of unemployment, known as U-6 (or the ‘underemployment rate’), captures a similar concept to the combination of ‘unemployed’ and discouraged workers described above. While U-3 is the official ‘unemployment’ rate, U-6 also includes: discouraged workers, all other marginally attached workers, and those workers who are part-time purely for economic reasons.

Figure 21 shows that, while the ‘unemployment’ rate (U-3) is now back at its 2007 lows (and only 50bps above its 2000 low), the U-6 underemployment rate remains 50bps above its 2007 level (and 1.6% above 2000). This suggests that plenty of slack remains before the US economy reaches full capacity. And the recent acceleration in the U-6 decline only further substantiates this.

Figure 21: Underemployment is a long way from its 2000 lows

Sources: Record Currency Management, Bloomberg, U.S. Bureau of Labor Statistics. To April 2016.

Taking the argument to the extreme, if we look to the lowest U-3 rate in history (2.5% in 1953) we can estimate a U-6 rate as low as 4% at that time (extrapolated based on a 1.6 ratio, due to lack of historical data). If “full” employment could, hypothetically, mean levels as low as the 1950s, there could be up to 4.5% of remaining slack.

…and the rest

As mentioned above, while it is plausible that a small potential supply of labour could be additionally found in those groups that are currently in education or disabled, it is fairly unlikely. No matter how deep we dig, there really aren’t many more people out there to put back to work!

Even if, under a very optimistic best-case scenario, an additional 10bp increase in the participation rate was generated from reducing the adult education and training rate, and an additional 20bp from bringing disabled and ill people back to work, this still only implies a maximum possible increase from these groups of around 750,000 people.

If a momentous childcare-subsidy programme was put in place, perhaps another 1 million or so could also be added; but, again, the prospects of this sort of cultural shift are very hard to measure.

Reality check: three million, not one hundred

So how many of those 100 million could really go back to work? Although the ‘real’ rate of unemployment is running at around 8.6% (U-6) – or 14 million people – this can never be brought down to zero, in any economy. Using the numbers calculated throughout this paper, Figure 22 summarises some scenarios that could be targeted in the U.S. for increased employment numbers, from the plausible to the not-so-plausible.

Figure 22: 2.6 million on the side-lines, not one hundred

Sources: Record Currency Management, Bloomberg, U.S. Bureau of Labor Statistics, Federal Reserve Bank of Atlanta

Without knowing the true natural (or equilibrium) rate of unemployment, we cannot know for certain how hot the U.S. economy would need to run in order to achieve any of the increased employment numbers in Figure 22. But adding back anything more than 2.6 million extra workers in the next few years (from those not currently employed) would appear to be a highly optimistic target. Even the 2.6 million target may be plausible but is still ambitious; this would mean driving the U-6 underemployment rate back to levels last seen in 2000, at the height of the Dot-com bubble, and could take years to achieve. Moreover, the Trump administration will also have to contend with longer-term demographic trends, which are shifting more and more of the American population into retirement. In short, no matter how aggressive Trump’s stimulus policies, getting even 3 million people “back to work” will be a tall order.



[1] Aaronson, D., Davis, J. and Hu, L. Explaining the Decline in the U.S. Labor Force Participation Rate. Chicago Fed Letter, No. 296, March 2012. https://www.chicagofed.org/publications/chicago-fed-letter/2012/march-296

[2] Kashkari, N. Why I Dissented. Federal Reserve Bank of Minneapolis, March 2017. https://www.minneapolisfed.org/news-and-events/messages/why-i-dissented

[3] Federal Reserve Bank of Atlanta, 2017. Labor Force Participation Dynamics. https://www.frbatlanta.org/chcs/labor-force-participation-dynamics.aspx

Where are we in the dollar cycle?

  • The US dollar is at an inflection point. Can the dollar’s cycle continue in the face of convergent economic fundamentals and central bank coordination?

US policy rate normalisation does not guarantee dollar strength – real interest rate support crucial

The US dollar is at an inflection point. On one hand the real exchange rate is yet to comfortably pass its long run average, implying further room for appreciation, while on the other the humdrum chorus of gloomy global economics and a false hopes for US rate normalisation could threaten to pour water on the fire. Indeed, Fed watching has become a favourite pastime for many, presumably in the hope of uncovering clues as to the dollar’s next step, however focusing on the Fed alone is unlikely to uncover the dollar’s secrets. Unfortunately, Fed normalisation cycles alone have never truly been enough to predict a sustained appreciation in the US dollar, and there is no discernible relationship between US rate normalisation, and the nominal or real exchange rate (figure 1). This is natural as currency is and always will be a relative concept.

With this in mind, and given small differences in the level of development between the US and most other developed markets, real interest rate differentials can help us to determine the dollar’s true attractiveness and in turn explain a good deal of behaviour in the medium run (in fact, interest rates and price differentials form the basis for most currency valuation models).  Increasing real interest rates (ceteris paribus, of course) help drive the dollar higher but this effect is more prominent in longer-dated real interest rates than shorted dated. The message one could take is that short term real interest rate normalisation is a necessary but not necessarily sufficient condition for dollar strength, and the US dollar’s destiny may be determined more by elusive long-run forces, in other words natural real interest rates.



Figure 1 and figure 2. Source: Macrobond, J.P. Morgan

Is there scope for real rate divergence?

In view of the theory that US dollar appreciation requires an ascension of US real interest rates relative to the rest of the world, the US dollar cycle may have hit a stumbling block. Declining natural real interest rates have been a global phenomenon (the BIS here and BoE  here have written extensively on this topic), with lower growth, a savings glut and rising inequality as cited culprits. For significant divergence in the long-end of the curve, these fundamentals in the US need to improve. This will be a challenge; even the Federal Reserve has been forced into submission by the secular stagnation thesis as the bank’s long-run projections of interest rates have continued to shift towards that of the market. A similar reality is also evident in the short end of the curve; forward guidance, global economic coordination and increased responsiveness to shocks by central banks are creating fewer economic surprises, fewer opportunities for divergence, and are suppressing interest rate and currency volatility (figure 3).


Figure 3. Source: Bloomberg, Macrobond, Citi Economic Surprise Index

Can the world cope with a stronger dollar?

While the bar for a stronger dollar would seem high, we believe the US can cope with a stronger dollar, and economic impacts should not hold the cycle back. The US economy is relatively closed with trade making up only 30% of GDP, therefore a stronger dollar is unlikely to pose a large threat to growth via the trade route. Additionally, we should not be too concerned about the currency’s impact on inflation in the US; currency pass-through to inflation is  the lowest among G10 economies (figure 4), in part due to the limited degree of openness but also because the vast majority of US imports are denominated in dollars, limiting the impact on inflation to volume effects only (as foreign imports are diverted elsewhere). 

We think that the rest of the world can too cope with and may indeed benefit from a stronger dollar. Firstly, most developed economies are not troubled by high inflation and would likely welcome a boost in price levels. In particular, the Bank of Japan and European Central Bank are plagued by currency strength and diminishing returns to ever expanding easing programs. In the Emerging Markets, a stronger dollar is not quite what it used to be; most sovereigns have reduced hard currency borrowing significantly, and while in some countries corporates may still be exposed to fx liability risk, EM central banks boast larger foreign currency reserves than in previous hiking cycles, and governors that are more adequately equipped skill-wise to deal with all the economic shocks that a stronger dollar would bring. 


Figure 4. Source: “The International Price System” , Gita Gopinath (2015). Eurozone is an average of Italy, France and Germany

What next if negative rates fail

  • In response to almost a decade of QE and with little discernible effect, central bankers have now resorted to negative interest rates. What is the zero lower bound and are below zero rates having the desired effect?
  • If negative interest rates fail to halt deflationary momentum, could more extreme options such as ‘helicopter money’ be a viable next step?

Negative interest rates can now be seen across the developed world, ranging from -0.1% in Japan to -0.75% in Switzerland. As more central banks dive into negative territory and with other such as the BoE not ruling the option out, it raises questions such as how low can negative rates go, and are negative rates effective in stimulating demand? In answer to the first question, theory goes that since cash holds an implicit carry of zero percent, the zero lower negative bound should be equal to the cost of insuring and storing the cash. In reality, the zero lower bound may be unlimited. If banks were to create additional charges for withdrawing cash exactly proportional to the negative interest rate on offer, the monetary benefit of withdrawal would be zero. To make things even more difficult for cash hoarders, in years to come there may be no cash to hoard; on the 15th February the ECB council voted to scrap 500 Euro notes, which represent around 30% of notes in circulation. In a cashless economy there is technically no lower bound.

Policy Rates

It would appear that whether negative rates work is directly related to the ability of savers to withdraw their deposits. In a world where there are no barriers to deposit withdrawal, negative interest rates could reduce the pool of loanable funds in the banking system and push up nominal interest rates. Of course, central bankers understand this, which is why we can expect to see further attempts to move towards a cashless economy. Already, negative interest rates are having unintended consequences. For example, most banks affected are showing reluctance and or the inability to pass on negative rates to depositors, while flatter yield curves from central bank bond purchases erode earnings on longer duration assets, hitting the bottom line through net interest income. When QE and negative interest rates are supposed to spur credit creation and animal spirits, fragile banks profitability and rising risk premiums will not help the cause. The Bank of Japan in its most recent move argued that the channel by which negative interest rates work is through asset reallocation and inflation-friendly exchange rate depreciation. The logic is correct, but currency is ultimately a zero sum game, and with disinflation becoming a global epidemic, a beggar thy neighbor currency policy is destined to fail, unless others (the US) can successfully shoulder massive  currency appreciation.

If negative rates prove problematic, where will the next round of easing come from should we face another global downturn? One potential policy gaining traction in the financial world is ‘Helicopter Money’, a term coined by the economist Milton Friedman. Conventional QE as we know it, indirectly reduces government borrowing costs and is not permanent (newly created reserves eventually disappear as debt matures).  Helicopter money in a modern format would be permanent QE, directly financing government spending or tax cuts indefinitely by purchasing bonds specifically issued by the government, either in the primary market (unlikely) or through the secondary market (legally and politically more defensible). This process could provide the boost the world economy needs, but there are dangers that need to be considered before a central bank and government decide to take the plunge.

For helicopter money to work , firms and households need to believe that any extra income received is permanent and will not remain a liability to the treasury, to be funded by tax at a later date. This can only be achieved if the bonds are wiped from the central bank’s balance sheet altogether, forgoing future payment. This free money would be appear fantastic to politicians and budget setters, but the central bank would find itself insolvent. While insolvency is not technically an issue at a central bank, the new structure of the balance sheet means its interest paid on excess reserves would likely exceed interest received on its assets. In this case more reserves are required to cover any income shortfalls, the balance sheet would grow out of control, and the central bank inevitably loses control of inflation. The alternative is to keep interest rates low enough such that interest expense matches asset income and the bank continues business as usual. This however implies maintaining low interest rates for an extended period, and again would probably result in loss of control over inflation.

Looking at the bigger picture, we cannot escape the truth that we are in an new era of poor demographics and subdued productivity growth, where the need to de-lever private and public balance sheets hinders the ability to generate growth and inflation. The longer central banks wholly bear the burden, the more they will be pushed into a corner with few remaining options. For these reasons, helicopter money or other extreme options such as wide-scale debt write-downs may gain more support across the board, with the understanding that a reset or purge of some sort may be a painful but necessary course of action in the long run. In some sense it has already begun; when the Bank of Japan owns over 30% of government T-bills and bonds, with little hope of reversing course, can we call it helicopter money yet?


Labour Share of GDP and the Balassa-Samuelson Effect

  • The labour share of GDP appears to be an important catalyst in the workings of the Balassa-Samuelson effect.
  • Understanding it can allow us to better infer currency valuations and the rate at which productivity growth translates into exchange rate appreciation.

The Balassa Samuelson effect postulates that productivity varies more by country in the tradable goods sector than in the non-tradable sector. This is on account of the inter-country competitive effects in the tradable sector and higher labour intensity of goods / services provided by the non-tradable sector. For example, there is only a finite number of haircuts a hairdresser can provide regardless of whether he or she is in Zurich or in Dhaka. Additionally, haircuts in Zurich cannot be easily arbitraged with haircuts in Dhaka regardless of the price differential.

The fact that tradable sector productivity drives income differentials however also has the effect of driving differences in real exchange rates. This operates through the labour market channel whereby wages in the non-tradable sector of countries with productive tradable sectors tend to get bid up given some assumption of a unified intra-country labour market. The lack of arbitrage across countries’ non-tradable sectors means that these higher price levels may be sustained without the need for any equilibrating factors. In sum, more productive economies can sustain higher real exchange rates over the long term structurally and this is why Zurich may feel “expensive” and Dhaka may feel “cheap” to a tourist.

While the theory above is well known, studied and empirically tested in the academic literature, something looked at less often are the factors that drive this effect. Something that might be important and is investigated not as often is the labour share of GDP (wage compensation as a share of GDP). Let us take the example of two countries that have had relatively similar levels of GDP per capita (measured in USD terms) over time; Switzerland and Qatar. For broadly similar levels of per capita productivity, the chart below shows how Switzerland can be seen to be noticeably more “expensive” than Qatar. Here, we measure how expensive a country might be by its real exchange rate, which is the ratio of its PPP exchange rate (per USD) to its actual exchange rate. For countries that are more expensive than the US, that ratio is greater than 1 and the opposite is true for those countries whose ratio is below 1. Each dot represents one year of the Swiss and Qatari real exchange rates from 1980 to 2015.

Figure 1 : Source IMF World Economic Outlook, October 2015

Figure 1 : Source IMF World Economic Outlook, October 2015

As is evident, there is a large gap between the two countries’ price levels for similar levels of per capita income. One possible explanation of this might be the fact that a much smaller fraction of Qatari GDP goes to labour relative to Switzerland. According to the latest estimates from the Penn World Tables, around 18 % of Qatar’s GDP accrues as wage compensation to labour while the number for Switzerland is 70 %. These lower levels of “pass through” from GDP to labour might explain why non-tradable price levels in Qatar continue to remain well below those in Switzerland. This is because a majority of the oil rents accrue to a select number of capital owners / the royal family and there is thus a limited wage pull effect in the non-tradable sector. This effectively constrains any significant real exchange rate appreciation as well as the underlying price level. The chart below highlights these differences across some major economies :

Figure 2 : Source Penn World Tables, 2011

Figure 2 : Source Penn World Tables, 2011

What can be taken away from this analysis is that some attention should be paid to currency pairs where each of the legs have drastically different levels of labour shares of national income. For such currency pairs, the dynamics surrounding reversion to PPP price levels or the impact of productivity growth on real exchange rates may be more complicated than what textbook economics might say. In the case of Qatar and Switzerland, it is evident that there are numerous other factors driving their exchange rate valuations including Switzerland’s status as a safe haven currency, the impact of the Qatari Riyal peg and oil prices, and an almost endless supply of low-skilled / low cost South Asian labour that the Gulf countries have managed to attract over the last few decades. These considerations aside, we think that the labour share of GDP is one additional useful metric to think about when assessing the impact of productivity growth on exchange rate valuations via the Balassa-Samuelson effect.