French household consumption in 2019: weak and strong at the same time

Has household consumption, the driving force behind French growth, stalled? Or was it actually in the process of rebounding? In 2019, household consumption rose at an average annual rate of 1.5% in real terms, which is considered to be a disappointing performance. But “disappointing” on what grounds and from which standpoint? Are we really dealing with a feeble rebound? These are difficult questions to answer, since everything depends on the perspective we take and the determinants we look at. In this article, we will try to put household consumption into context, and provide answers and explanations for the above issues. In a descriptive analysis in part one, we examine household consumption’s role as a growth engine, its momentum and composition. The second part is explanatory. Our analysis focuses on household consumption’s (lack of) momentum since 2008 in general, and in 2019 in particular.

The initial impression that French households were especially thrifty in 2019 has not really been confirmed. We see it more as a mixed performance, with both positive and negative factors. Rather than a stalled growth engine, it looks more like household consumption is simply restricted. Indeed, it benefits from a full fuel tank, thanks to large purchasing power gains, a sharp decline in the unemployment rate, ultra-low interest rates, dynamic lending, positive wealth effects and renewed confidence. Yet other factors are also hampering household consumption: the persistent impact of the lost decade of feeble purchasing power gains in 2008-2018, the composition effects of the purchasing power gains, the increased weight of pre-committed expenditures, the relatively low level of household confidence, and the complex articulation of the many and varied fiscal and budgetary measures, some of which supported purchasing power while others did not. Time is another factor that we must add to this list: consumption does not respond immediately and fully to purchasing power gains, as strong as they may be. Using rugby as a metaphor, if France scored a try in 2019, will it be awarded a conversion in 2020? Unfortunately, we will never know whether the rebound in household consumption would have been amplified as expected, because the outbreak of the Covid-19 pandemic has changed everything.

Household consumption as a growth engine: a look at the dashboard, its weighting, pace and composition

Orders of magnitude

We will begin by reviewing a few orders of magnitude and elements of language. Why, for example, do we say that household consumption is the growth engine of the French economy? It’s a matter of its weighting and growth rate. Household consumption now accounts for a little more than 50% of French GDP (down from 60% sixty years ago), which makes it by far the most important component of economic growth (see chart 1). Compared with other countries, however, France is not the country with the highest consumption share. That record is held by the United States, where household spending accounts for 66% of GDP (see chart 2).

It is the combination of its heavy weighting as a share of GDP plus its rapid growth rate that makes household consumption the component that contributes most to French GDP growth: it has contributed a little more than half on average since 1950, compared to only about a quarter for public consumption, investment and exports. Yet this contribution fluctuated over time and over business cycle (see chart 3): it was especially high during the second half of the 1970s (contributing 60% of GDP growth) and very low during the 2009-2011 mini-cycle (33%).

During the most recent business cycle (2013-2019), household consumption contributed half of GDP growth, which is in line with the historical average. What is interesting is that it was followed closely by investment.

Over the three most recent years (2017 to 2019), the average annual contribution of investment was even 0.2 points higher than household consumption (0.9 points and 0.7 points, respectively). This is rather rare and it fuelled debates on the sluggishness of consumption. It is worth keeping in mind, too, that household consumption not only serves as a growth engine during expansion phases, but also acts as a buffer during recessions. Indeed, it rarely declines and any fluctuations are much smaller than for investment or exports.

The strength of the 2019 rebound in household consumption: a matter of perspective

In 2019, French household consumption rose at an average annual rate of 1.5% in volume, after 0.8% in 2018¹. This is a tangible rebound, with consumption rising nearly twice as fast as in 2018. More often than not, however, it is qualified as “disappointing” or a “missed opportunity”. Why this disappointment? First of all, it is disappointing relatively to the purchasing power gains reported in 2019 (see chart 4): the increase in consumption pales when compared to the 2.1% increase in purchasing power (this point will be discussed in greater detail in the beginning of part two).

Once put into perspective, however, there is nothing unusual about this gap between the two variables, to the contrary. And the acceleration in household consumption between 2018 and 2019 (+0.7 points) was perfectly acceptable compared to that of purchasing power (+0.9 points). Yet there is something misleading about the acceleration of consumption: it was facilitated by a particularly sluggish performance in 2018. Moreover, with the dissipation of the temporary factors which hampered consumption in 2018² and the vigour of purchasing power gains, the conditions seemed to have come together for a major rebound in household consumption in 2019. Yet it only matched the growth rates of the three previous years, which were by the way extremely regular (1.5% in 2015, 1.6% in 2016, and 1.7% in 2017, or an average of 1.6%).

Household consumption also seems to be sluggish when compared with its historical average growth rate of 3.1% a year since 1950. chart 5-A clearly shows when household consumption has outperformed or underperformed its long-term trend. Yet the pertinence of this comparison is open to question given how much the world has changed in the meantime. The average has an upside bias due to the rapid pace of consumption in the 1950s and 1960s (5.1%) as well as in the 1970s (4.1%) (see chart 5-B). Adjusted for population growth to take into account demographic trends, these average growth rates were not as high (by about 1 point), but are still robust: 2.4% since 1950, 4.1% in the 1950s and 1960s, and 3.5% in the 1970s.

When calculated for a more recent period, starting in 1980, the average slips to only 1.7% (1.2% for per capita consumption). Seen in this light, the 2019 performance of 1.5% (1.4% per capita) is not particularly low and close to the average. Yet this more recent average has the opposite shortcoming of the other longer term average: it has a downside bias due to the slower growth rates since 2008. This year marks another rupture, after the trend break in the late 1970s and early 1980s. From 1980 to 2007, household consumption grew at an average annual rate of 2.1% (1.6% per capita), before declining to 1% over the period 2008-2019 (0.5% per capita). Chart 5-C also illustrates the slowdown in household consumption over the course of various business cycles.

We can provide an even clearer illustration of the trend break in 2008-2009: chart 5-D is the same as 5A, although we have zoomed in on the period since 1995. The virtual stagnation of consumption between 2008 and 2012, when growth averaged 0.5% a year, bears the scars of the double shock of the Great Recession and the European sovereign debt crisis. Seen in this light, there is something positive about the quasi-stagnation of consumption, which displayed resilience rather than sluggishness. Since 2013, the slope of the curve has picked up again, but only slightly (to an average annual rate of 1.2%). The year 2019 fits within this milder trend.

A final touch is to be added to this picture. A comparison with the Eurozone shows that France has lost its over performer position: between 2014 and 2018, French private consumption³ grew at a slower pace than the Eurozone average, with a negative differential of 0.3 points on average, whereas the differential had favoured France since 1998, by an average of 0.7 points (see chart 6). Yet the negative differential for France is not uniquely a sign of relative weakness: the Eurozone average has been strengthened by a catching-up effect in the countries hit hardest during the European sovereign debt crisis, at a time when German consumption was also going strong. Under these conditions, it is harder for France to remain above the average. All in all, 2019 can be seen as a good year for France since it swung slightly above the Eurozone average again (+0.2 points).

To summarize, it has been well established that French household consumption has lost momentum over the years. This leaves the question of its dynamics in 2019 and how best to gauge it. Did consumption really lack vigour? Which benchmark is best for comparing the 2019 growth rate? The 2.1% average for the period 1980 to 2007? This was a more stable period pre-dating the financial shock and Great Recession of 2008-2009. Using this benchmark, consumption underperformed in 2019. Or the 1% average of 2008 to 2019? This was a more turbulent period encompassing two crises, but it is also more recent. Compared to this period, 2019 seems like a rather good year, or at least a recovery year. Clearly, the labels “sluggish” or “subpar” that have been pinned to consumption in 2019 are perfectly relative and depend on the perspective that is used.

In our point of view, French household consumption has entered a new, slower growth regime since 2008, with an average annual growth rate of between 1% and 2%. Our previous benchmark of an average growth rate of 2% is now at the upper end of the range. Seen in this light, the year 2019 was a perfect transition year, in which household consumption was strong and weak at the same time.

Structure of consumption: household arbitrable spending has been picking up since 2014

The scope of our analysis covers final household consumption expenditure, i.e. spending that is directly incurred and payable by households. To analyse the structure of consumption, however, we must take into account another concept, effective consumption, which is higher, because it also integrates expenditure covered by administrations that directly benefit households (see chart 7). Effective consumption is the sum of final household consumption expenditure plus so-called “individualisable” expenditure by public administrations (education, healthcare, social welfare and housing) and by non-profit institutions servicing households (NPISH). The weight of these “social welfare” expenditures as a share of effective consumption has increased significantly over time, rising from 15% in 1959 to 25% in 2019, reflecting the expansion of the French welfare state⁴.

The structure of effective consumption by French households has evolved enormously since 1960 (see chart 8). Some of the main changes are a big decrease in the weighting of food and clothing, and a big increase in the weighting of housing (which is now by far the biggest category of consumption), healthcare and, to a lesser extent, communications. The size of these transformations vary depending on whether we look at data in value or volume terms, but there is no fundamental difference in the overall message⁵.

To analyse the change in the structure of consumption that interests us most in this article, we must distinguish between fixed and arbitrable expenditures, depending on how easily they can be compressed. Fixed expenditure comprises so-called pre-committed spending and expenditure that is not easily compressed.

Households have little or no control over these expenditures. The INSEE defines pre-committed expenditure as spending that is “realised within the framework of a contract that cannot be easily renegotiated in the short term.” They comprise spending on housing (including imputed rent as well as spending on water, gas, electricity and fuel used in homes); telecommunication services, cafeteria meal plans, television services (television charges, pay channel subscription fees), insurance (excluding life insurance) and financial services.

Expenditures that are not easily compressed cover essential needs, including food and non-alcoholic beverages, healthcare, fuel and lubricants, transport services and education.

Arbitrable expenditures, which can be considered discretionary purchases “for pleasure”, include items that can be easily compressed: clothing and footwear; furniture, household cleaning and maintenance products; leisure and culture (excluding television services); hotels, coffee shops and restaurants (excluding cafeteria meal plans); alcoholic beverages and tobacco products; other goods and services (except insurance and financial services – with the exception of life insurance); and communication products (excluding telecommunication services).

Sixty years ago, in 1959, the final consumption expenditure of French households could be broken down into quasi-equal parts between fixed (49%) and arbitrable (51%) expenditures. By 2019, the situation had changed dramatically with the weighting of the first rising to 57% and the second dropping to 43% (see chart 9). This shift in weightings have significant consequences, as we will see below. The big increase in the weighting of fixed expenditure is due to pre-committed expenditure (+19 points), especially imputed rent (+11 points). The weighting of expenditures that are not easily compressed has declined even more than that of arbitrable spending (by 11 points and 9 points, respectively).

The share of pre-committed expenditure increased mainly during the first half of the period, from 15% in 1959 to 32% in 1993. It then levelled off for the next decade before picking up mildly, and temporarily, thereafter. It peaked at 35% in 2013 before declining slightly to 34% in 2019.

Looking back at the slowdown in household consumption over time, and especially since 2008, we can see that it generally hit all spending categories (see table 1). Taking into account the weightings of the various categories, it is transport and leisure that made the biggest contributions to the slowdown (accounting for about a quarter), followed by housing (for a little over 15%), healthcare and other goods and services (about 10%). The scars of the 2008 crisis can be seen in the slowdown in purchases of durable goods (automobiles, furniture), semi-durable goods (clothing) and leisure, which are all arbitrable categories that do not qualify as basic needs, and that can be easily cut or postponed⁶.

The slight upturn in consumption during the recent period (+0.5 points in 2013-2019) was not widespread, but touched on the categories that had been curbed the most previously, with the exception of communications services (which continue to slow). Spending on education also contributed to the rebound, while the other categories continued to slow (food, beverages, tobacco products, housing, healthcare and other goods and services).

Lastly, looking solely at the year 2019, there was an increase in consumption virtually across the board, with the exception of food, beverages and tobacco products (for tobacco products, this can be attributed to the health ministry’s decision to raise cigarette prices sharply in the fight against smoking).

If we now look at the issue in terms of fixed and arbitrable expenditures (see chart 10-A), we can clearly see the impact of the 2008 crisis in the decline in arbitrable categories. Chart 10-B illustrates well the adjustable nature of these expenses: arbitrable expenses fluctuate much more than the other two categories. They are basically more cyclical. The good news is that they have picked up since 2014, signalling less restrained consumption. And when looking at the question of whether household consumption was strong or weak in 2019, this trend clearly tilts the balance towards strong growth.

Restrained consumption

Our descriptive analysis in part one concludes that French household consumption grew at a fair pace in 2019: in the light of recent trends and its composition, it was not all that weak. As to its key determinants, it is harder to judge whether the growth of household consumption was in line or not due to overlapping positive and negative effects and a mix of tailwinds and headwinds. Econometric modelling could be used to quantify the residue, to determine whether the unexplained part is important or not. Without it, we will simply review a number of these determinants, along the following explanations: 2008-2018, a lost decade of feeble purchasing power gains; the increased weight of pre-committed expenditure; the composition of purchasing power gains; the puzzling insensitivity to the unemployment rate fall; the likely existence of Ricardian effects; the relatively low level of confidence; low interest rates and limited wealth effects.

Before going any further, we must keep in mind that household consumption can also be subject to wide variations from one quarter to the next, that have nothing to do with these determinants, due to exogenous shocks such as unusually harsh or mild weather conditions (which drive up or down energy expenditure), the introduction of “vehicle scrapping premiums,” changes in VAT rates, strikes and labour unrest, and major sporting events. These infra-annual disruptions carry over to average annual growth rates, artificially inflating or lowering them. This distorts our fundamental analysis, by signalling a strong or weak performance where there might not necessarily have been one.

Feeble purchasing power gains: 2008-2018, the lost decade

Purchasing power trends are one of the main explanations for the slowdown in household consumption since 2008. Before going any further, we must first define what we mean by “purchasing power”. This term is widely used but does not always mean the same thing for everybody. It can be used to cover very different individual situations that deviate from the macroeconomic benchmark defined by the national accounts. This is one reason for the sometimes big gap between people’s feelings and experience and the aggregated figures, which can interfere with their behaviour.

Using the national accounts definition, purchasing power is a measure of real gross disposable income (GDI):

Disposable income refers to the portion of household income that is available after paying taxes and social contributions (whose weighting is around 30% of GDI).

Pre-tax income is comprised essentially of earned income (including wages and the income of self-employed individuals), which accounted for 57% of pre-tax income and 72% of GDI in 2019, as well as social benefits (27% of pretax income and 35% of GDI) and capital income (15% of pretax income and 19% of GDI).

Chart 11 shows the long-term changes in the composition of GDI, especially its growing “socialisation” (mirrored in household consumption) via the trend parallel increase in the weighting of social welfare benefits and taxes.

Gross income refers to the fact that it is not adjusted for fixed capital consumption linked to self-employed businesses and home ownership. It is real once it has been adjusted for prices.

Purchasing power gains are thus the difference between the nominal GDI growth rate and inflation⁷. According to this definition, declines in purchasing power at the macroeconomic level are extremely rare and small in scope (see chart 12). The latest decline was fairly recent and lasted for two years (-0.4% in 2012 and -1.2% in 2013), which left its mark on French households.

Chart 12 also shows the factors behind the momentum of purchasing power gains, and especially, for the part that interests us in this study, the reason of their weakness since 2008. Two trends are particularly striking: the net decline in inflation, which is a positive factor, and the net decline in earned income growth, which reflects tough job market conditions and is obviously a negative factor. Since 2008, purchasing power has risen at an average annual rate of only 0.9%, compared to a trend of 2.4% since the mid-1980s, and 3.3% if we look as far back as the 1960s.

The sluggish pace of gains since 2008 is one of the reasons why purchasing power is a recurrent focal point of debate. If we look at purchasing power in terms of consumption units (CU), it is even easier to understand why it is such a big issue today. Consumption units allow us to take into account demographic trends (number and composition of households), which gives us a better picture of individual perceptions of purchasing power gains. Measured by CU, declines in purchasing power occur more frequently. Since 2008, purchasing power gains have shrunk to an average annual rate of only 0.3%. In 2018, purchasing power by CU was barely higher than the 2008 level (see chart 13). Clearly, we can speak of a lost decade.

Arbitrable disposable income, an indicator that deducts pre-committed expenditures from disposable income, provides an even better idea of how households are feeling about their personal wealth. This is the truly disposable part of their income, which households can use to pay for other spending. The impact of the pre-committed expenditure bigger share of household budgets can be seen in the slight uncoupling between arbitrable purchasing power by CU and purchasing power by CU between the late 1970s and the first half of the 1980s. This gap did not widen thereafter. Since 1986, the two indicators have moved more or less in tandem, at an average annual rate of 1.2%.

Households could have offset the impact of sluggish purchasing power gains by dipping into their savings to maintain the pre-2008 growth rate of consumption, or at least to limit the slowdown. Yet this was not the case. The growth of household consumption dipped in line with the slowdown in purchasing power gains, or even slightly more (see chart 14). We have highlighted the years 2008 and 2009 in chart 14 to show the impact these two crisis years have had on consumption and purchasing power, and to show how they have evolved in relationship to each other.

What is striking is the big split in 2009 between the rebound in purchasing power (due to lower prices that year) and the downturn in consumption.

The period 2008-2012 also stands apart because it was marked by two crises and an aborted mini-recovery between the two. These difficult years exacerbated the low level of consumption and purchasing power gains, and are thus not representative. During these years, household consumption did not grow as fast as purchasing power gains (the 2009 effect), and this situation was reversed thereafter, with consumption slightly outpacing purchasing power gains in the period 2013-2019. Yet once again, averages can be misleading. This positive spread can only be seen in three of the seven years in question (2013, 2015 and 2016). Moreover, it is biased by the year 2013, when purchasing power declined sharply, but not consumption. In 2014, 2017, 2018 and 2019, consumption did not grow as fast as purchasing power. The year 2019 does not seem to be an exception, but rather the rule.

However, the year 2019 stands out from the rest because purchasing power gains reported that year had a broader base, buoyed by improvements in a greater number of components (slightly stronger earned income thanks to a healthier job market situation, more social welfare benefits and less fiscal pressure thanks to government measures, and lower inflation). This broad-based gain fuelled expectations that household consumption would increase in line with purchasing power, at a rate of close to 2%. Yet the increase in consumption was limited to 1.5%, which contributed to a feeling of disappointment.

Yet we must also take time into account, among other factors. Consumption does not respond instantaneously and fully to purchasing power changes. The years 2009 and 2013 are good examples. According to currently accepted estimates, a 1% increase in purchasing power will eventually trigger a 1% increase in consumption in the long term, but in the short term, half of the extra income is spent and the other half is saved. Seen in this light, there is no reason to be disappointed by the pace of consumption in 2019. It correctly responded to the increase in purchasing power. As this time factor was expected to dissipate, 2020 was anticipated to be the year of the “true” rebound in consumption, but with the outbreak of the Covid-19 crisis, we will never know whether that would have happened.

The weight of pre-committed expenditure

The lag in the response of household consumption to changes in purchasing power is perfectly normal: it takes time for households to realise their situation has improved, to measure its scope and to be reassured of its permanence. The delayed reaction can also vary depending on the level of confidence, income constraints and the structure of consumption.

Greater attention is precisely being paid to the structure of consumption due to the growing weight of pre-committed expenditure (see above). Its weighting contribute in part to the negative feeling concerning standards of living in France, the gap between purchasing power statistics (rarely negative) and surveys of the French people (who give the opposite impression), reflecting the reduction in financial leeway due to spending constraints. And this negative feeling can contribute to the fact that major macroeconomic purchasing power gains do not boost household confidence and consumption all that much. Income restraints seem to have been lifted at the aggregate level, but are not perceived as such, or insufficiently so, at the individual level.

Moreover, Beatriz, Laboureau and Billot (2019)⁸ show that in the short term, the fixed expenditures are not very sensitive or even insensitive to purchasing power gains. They cross these results with the standard of living of households, knowing that low-income households have more expenditures that are not arbitrable (see charts 15 and 16). The spending that is most sensitive to purchasing power gains are arbitrable expenditures. On one hand, arbitrable spending represents a big share of consumption for the wealthiest households, who have the lowest propensity to consume. But on the other, for arbitrable expenditures or those that are difficult to compress, household categories with lower incomes are more likely to consume the increase in revenue. Put together, this is likely to extend the transmission period for household purchasing power gains to carry over to consumption.

Faure, Soual and Kerdrain (2012)⁹ make estimates of income elasticity by product category, which provides quantified information that builds on the above sensitivity analysis: the weaker the income elasticity, the more constrained the spending. In these product categories, we clearly find spending that is difficult to compress as well as purchases with few constraints (high income elasticity), which correspond to arbitrable expenditure (see table 2). In general, spending on goods seems to have less income sensitivity than expenditure on services, but price sensitivity is similar for the two types of spending. Yet the first have slowed just as much as the second since 2008 (and not less), a paradox that can be attributed to the fact that the prices of goods grew relatively more rapidly than service prices (a reversal of the configuration that prevailed prior to 2008).

A study by Lelièvre and Rémila (2018)¹⁰ on the weighting of pre-committed expenditure provides additional information on how an unexpected inflow of money would be used based on standard of living. The types of spending taken into consideration cover a broader scope, including building up savings and repaying loans (see table 3). It is interesting to note that the proportion of households who would use this hypothetical extra income in these ways does not depend on their standard of living: about 20% of households would save the extra money, and about 10% would use it to repay loans. These are non-negligible proportions, and both cases are negative factors for consumption. Differences in standards of living come into play for food, clothing and healthcare (the proportion decreases as the standard of living rises) as well as for housing, leisure and helping family and friends (the proportion increases as the standard of living rises).

Income composition effect

The composition of income and the nature of purchasing power gains also have an impact on the sensitivity of consumption’s reaction to purchasing power gains. Indeed, the propensity to consume is not the same depending on the type of revenue, because the composition of income and the propensity to consume differ depending on the standard of living (see table4)^{11 12}.

Yet Bardaji, Lequien and Poissonnier (2014)¹³ were unable to determine that income composition was a factor in the feeble increase in consumption between 2008 and 2013. According to the authors, the results obtained “partially reflect the heterogeneity of agents more than differences in the propensity to consumer depending on the type of revenue”. For the Bank of France (2018, 2020)¹⁴, in contrast, when the model of household consumption takes into account the composition of gross disposable income (GDI), the quality of the equation is clearly improved: changes in the savings rate since 2005 are reproduced more accurately. The following results are obtained:

“A shock to wage income or social benefits, all other sources of income being the same, is consumed almost entirely and the savings rate remains virtually unchanged”;

“Tax cuts are split equally between consumption and savings”;

“For net financial income, estimates suggest that the propensity to consume is close to zero, which means shocks to these revenues solely affect savings.”

The Bank of France draws our attention to the fact that these are “average” results, which can therefore be affected by the distribution of tax cuts by income levels¹⁵. The elimination of the housing tax for lower income households (in three phases in 2018, 2019 and 2020) and the increase in the in-work bonus in 2019 should benefit from a greater propensity to be consumed. To the contrary, the capital gains tax cut in 2018 should result in higher savings in the long term.

Seen from this angle, it is easier to understand why the household saving ratio increased in 2018 and 2019, and how and why some of the measures to boost purchasing power gains have ended into savings. Using the same logic, we could also reasonably expect January 2020 income tax cut for middle-class households to have a major positive effect on consumption, although we will never know its real impact since it was wiped out by the Covid-19 crisis.

The unemployment rate and savings rate: the relationship seems upside down

Theoretically, there is nothing ambiguous about unemployment’s impact on consumption. When unemployment has a significant impact in econometric models, it always comes with a negative sign (and when the variable is the savings rate, the sign is positive). Looking beyond the direct impact on current income (reflected in the income or purcha-

sing power variable), unemployment also has an influence on income expectations and uncertainty: higher unemployment reduces income expectations and increases uncertainty, resulting in the building-up of precautionary savings¹⁶.

In a counter-intuitive manner, however, we come to the opposite conclusion based on observations in charts 17 and 18. The supposedly positive correlation between the unemployment rate and the household saving ratio does not seem to be working. To the contrary, they seem to be doing just the opposite. In level terms, a structurally high savings rate does indeed coincide with a structurally high unemployment rate. In terms of variation, however, the correlation between the two indicators seems to be negative: when the unemployment rate rises, the household saving ratio falls.

More specifically, the correlation is inconstant. Sometimes the unemployment rate and the savings rate increase in step, as expected (as in 2009 in particular). But this joint movement, either upwards or downwards, is not systematic. It might just be a question of a temporal lag, with one advancing or the other lagging. A two-way causality does exist between the two variables: the usual causality is that an increase in the unemployment rate carries over to the savings rate (via precautionary savings), but there is also a reverse causality in which an increase in the savings rate provides an early signal of a cyclical downturn and precedes an increase in the unemployment rate. In practice, the relationship between the unemployment rate and the savings rate appears to be more complex and ambiguous than the theoretical model.

The year 2013 is the first recent counter-example in which the unemployment rate and the household saving ratio move in opposite directions, with the first rising and the second falling. Faced with a shock to their purchasing power that year, households either opted to or were forced to dip into their savings to smooth out and maintain consumption.

In 2018 and 2019, the configuration was the reverse of the one in 2013: the savings rate rose while the unemployment rate fell, a trend started in mid-2015. It is possible that the unemployment rate’s downward impact on the savings rate was more than offset by other factors at work that increased the savings rate. Another complementary interpretation to this enigmatic insensitivity of the savings rate to the decline in the unemployment rate is that this latter is not a good measure of income uncertainty, at least not in all circumstances. Uncertainty could have arisen from other, more diffused sources, such as the social malaise manifest during the “yellow vest” crisis in late 2018 and early 2019. The strikes against the systemic pension reform in late 2019 and early 2020 are another sign of discomfort. The general climate of uncertainty that has reigned in recent years must also be taken into account.

Are French households Ricardian?

Fiscal policy in recent years might also have driven up the savings rate of French households via two channels: as a source of uncertainty and due to likely Ricardian behaviour. To simplify, a household is Ricardian (as opposed to Keynesian) when, faced with a permanent tax cut (for instance), it increases savings (and not consumption) because today’s tax cut will implicitly increase the taxes that will have to be paid tomorrow. Like when faced with unemployment, households will set aside precautionary savings to cover future tax increases. As a result, the expected positive impact of a fiscal stimulus on growth is either diluted or annulled. Inversely, the Ricardian equivalence can have a positive impact when faced with the negative effects of fiscal consolidation¹⁷.

Among the studies we have read, research on French data leads to contrasting results: sometimes the existence of a Ricardian effect is rejected, in any case before the 2008-2009 crisis¹⁸, and sometimes it is confirmed¹⁹, but even then it is found to be small in size, acting solely on the short term, and showing sensitivity to the estimation period²⁰. For Bardaji et alii (2014), the Ricardian behaviour of households also depends on the size of variations in the fiscal balance: this estimate suggests the presence of non-linearity and a “stronger household reaction when public savings undergo major variations,” but that are not significant in the opposite case.

Based on these results, it is possible to consider that for Ricardian reasoning, some of the tax cuts and other fiscal and social measures designed to boost purchasing power in 2018-2019 did not stimulate consumption but rather savings. This analysis is made all the more complicated by that fact that other measures were introduced during the period, not only those to boost purchasing power. There were also some financing measures. What can we say then about their impact on consumption? Which is stronger, the positive impact of the Ricardian hypothesis or the negative impact of Keynesian assumptions? Based on household confidence and consumer trends in 2018, the reaction seemed to be more Keynesian. This suggests that there is some asymmetry in the reaction of households, who tend to react more negatively to unfavourable purchasing power changes than positively to a favourable change.

The fiscal policy pursued over the past two years was supposed to have a positive effect on consumption, but it may have been attenuated by the uncertainty and the confusion engendered by the timing, number, diversity and mix of measures. This reduced the readability and understanding of fiscal policy as well as, possibly, its credibility and effectiveness. The combination of austerity and stimulus measures, a strategy that has been followed since 2014, resulted in a definite but incremental reduction in the fiscal deficit. If there is a Ricardian effect then it is limited, in one direction or the other. Most importantly, despite all the efforts and the reality of statistics, deficit reduction remains a subject of caution, and does not inspire confidence given the lack of visibility. Similarly, there is no obvious response to the questions “Will I benefit from a tax cut? And if yes, then how big, when and how long will it last?” Difficulties in understanding and believing policy measures, combined with the lack of certainty, clarity, and confidence, may have kept household consumption from responding more positively to measures to boost purchasing power.

The importance of confidence

Confidence is obviously a key factor in the consumption/savings trade-off that households must make when faced with changes in purchasing power. It informs us about “the state of mind of households and their propensity to consume” and enables us to capture some of the expectations not reflected in the other determinants of consumption²¹.

Taking a long-term perspective, such as that measured by the INSEE, a clear break in the level of confidence can be seen in 2008-2009 (see chart 19) that coincides with the trend break observed for consumption and purchasing power gains. Household confidence picked up strongly between 2012 and 2017, but remained relatively low compared to previous highs. Fluctuations in confidence can be linked to how the French view their own standard of living, such as that provided by the Crédoc survey of living conditions and aspirations. According to this survey, the French were effectively much more pessimistic in 2019 than in 1979. Forty years ago, 46% said that their personal standard of living had improved over the previous ten years. By 2019, this figure had fallen to 24%²². This “feeling” generally tends to follow purchasing power trends by consumption unit, and the good news is that it has been improving since 2017 (see chart 20).

In 2018 and 2019, household confidence was hit by major fluctuations. A series of shocks eroded confidence and consumption in 2018. It began with the fiscal measures introduced at the beginning of the year²³. The erosion continued with the transport strikes in Q2 2018. Confidence dropped again in Q3 due to energy inflation, which slashed purchasing power, as well as the announcement for 2019 of the partial deindexation of pensions, family allowances and housing subsidies. There were other sources of uncertainty including the introduction of a withholding tax system starting in January 2019 and the debates over pension and unemployment insurance reforms. The “yellow vest” movement also undermined household confidence in the last months of the year.

In 2019, household confidence resumed rising robustly, buoyed at last by the measures to boost purchasing power, ongoing job market improvements and weaker inflation. Between January and September, confidence improved over an exceptional series of nine consecutive months. By year-end 2019, it had regained the ground lost in 2018 and rose slightly above its benchmark average of 100. French households were showing cautious optimism. They were confident, but not overly so: consumption was in line with this uptrend, it was picking up and gaining momentum, but without sparks (see chart 21).

Weakness of interest rates and wealth effects

With low interest rates and bullish equity markets and house prices in recent years, it might seem surprising that French household consumption was not more robust. Concerning the wealth effect, the answer is easy for once. Recent research signals the existence of a wealth effect, but although it is econometrically significant and robust, it is still small in size²⁴. For Chauvin and Damette (2010), a 1% increase in aggregated wealth (financial and real estate) leads to an increase in household consumption of about 0.1%, or, using the marginal propensity to consume (MPC), a one euro increase in wealth leads to an increase in consumption of between 1 and 2 cents over the long term. According to the same authors, the financial wealth effect (of about 0.1% in elasticity, 4 centimes in MPC) is higher than the housing wealth effect (of about 0.06% in elasticity, 2 centimes in MPC) and the aggregated wealth effect is bigger in the short term (0.12% from a 2-quarter horizon) than in the long term (0.07%).

The interest rate effect, in contrast, is ambivalent, since two opposing forces are at work: a substitution effect (which leads to a negative correlation between changes in interest rates and consumption) and an income effect (positive correlation). In the case that interests us here, a decline in interest rates is therefore both a support factor (substitution effect) and a damper (income effect) for consumption (and vice versa for savings). In estimates using French data, the substitution effect tends to win out over the income effect. Yet we cannot rule out the possibility that the income effect has gained in importance and has had a more negative influence during this extended period of low interest rates. This reasoning is based on average figures, however, which masks the heterogeneity of household behaviour and situations. For example, the wealthiest households have more financial income than lower income households, but also a lower propensity to consume²⁵: for these households, and on the whole, the income effect might actually be rather low.

This is how Dossche et alii (2018)²⁶ defend the effectiveness of the ECB’s monetary policy. Looking at the change in interest paid and received by households in France, they observe a similar-sized decline in 2008-2017, so we can infer that the decline in interest rates had neither a positive nor a negative effect on gross disposal income, nor on consumption. Yet the authors claim that there actually was a positive effect because “lower interest rates have mainly redistributed resources from net savers to net borrowers. As net borrowers typically have a higher propensity to consume than net savers, this redistribution channel of lower interest rates supports aggregate consumption”.

****

To conclude, household consumption in 2019 can also be described as a thwarted rebound, one that stopped in mid-air, and we will never know whether it would have regained momentum as expected in 2020. Unfortunately, the massive recessionary shock triggered by the Covid-19 pandemic has changed everything. Chart 22 clearly shows the extent of the shock: in the two-month period of March and April 2020, household spending on goods plummeted and came back to the level of the late 1980s. This collapse was followed by a similarly spectacular rebound in May and June, when spending on goods recovered all of the ground lost, and even more, because in June, consumption was 2.7% higher than in February.

As encouraging as that may seem, this rebound was largely automatic and tells us nothing about what comes next, nor how strong growth will be once the catching-up effect has dissipated. This rebound must also be kept in perspective since it only concerns spending on goods (50% of final household consumption), which was not hit as hard by the health crisis and the implementation of distancing measures as the consumption of services (transport, hotels and restaurants, recreation and retailing, which account for 33% of final consumption), which will take longer to return to normal, if it ever does.

Will there be a before and after 2020, like there is a before and after 2008? What repercussions will the Covid-19 crisis have on consumption patterns and growth? The debate is open and anything is possible. Will households opt for buying locally, short supply chains, organic foods, local sourcing and e-commerce? Will there be fewer leisure activities, travel and trips abroad? Will there be a rupture, extension, adaptation or acceleration of existing trends? More savings and less consumption? Have we come to the end of the consumer society and mass consumption? Will this be the dawn of responsible consumption? All of these questions will need to be examined in future research.

Hélène Baudchon

¹ All of the data referred to in this article are taken from national accounts available as of mid-June 2020

² Two factors were predictable but highly disruptive (Q1 tax increases and the Q3 introduction of the new anti-pollution standard WLTP – Worldwide harmonized Light vehicles Test Procedures), while three others were unpredictable (mild winter weather conditions and transport strikes in Q2, and the yellow vest movement in Q4).

³ Exceptionally, for this example we use data for private consumption (household spending plus expenditure by NPISH - non-profit institutions serving households) since there are no Eurozone statistics for household spending alone, as used in the rest of this article. For France, there is no difference in terms of the growth rate.

⁴ For more details, see Joan Sanchez-Gonzalez (2016), In 2015, the community accounted for a quarter of household consumption, INSEE Première, n°1618, September

⁵ For more details, see: Fifty years of household consumption, INSEE Références collection, 2009 edition

⁶ For more details see: Gaëlle Gateaud, Sylvain Heck, Brigitte Larochette, Nathalie Morer, Joan Sanchez-Gonzalez, Philippe Serre and Thomas Veaulin (2015), Since the 2008 crisis, households have reduced purchases of goods and services that are easiest to cut or postpone, in L’Economie française, 2015 edition, INSEE Références.

⁷ Measured by the change in the personal consumption expenditure deflator, which is fairly close to that of the consumer price index.

⁸ Mikael Beatriz, Thomas Laboureau and Sylvain Billot, 2019, What is the link between household purchasing power and consumption in France today? An analysis by household category and type of consumption, Note de Conjoncture INSEE, June.

⁹ Marie-Emmanuelle Faure, Hélène Soual and Clovis Kerdrain (2012), Household consumption during the crisis, Note de Conjoncture INSEE, June

¹⁰ Michèle Lelièvre and Nathan Rémila (2018), Pre-committed expenditure: what share of household budgets?, Dossiers de la DREES, n°25, March

¹¹ The propensity to consume decreases along with the standard of living.

¹² See research by Xavier Bonnet and Hélène Poncet (2004), Income structures and differences in the propensity to consume, Working document, INSEE, December

¹³ José Bardaji, Matthieu Lequien, Aurélien Poissonnier (2014), French household consumption since 2009: role of the fiscal/social system, in L’économie française, 2014 edition, INSEE Références

¹⁴ Bank of France (2018), Macroeconomic projections for France, box “The composition of household income, the household saving ratio and household consumption”, September, and full article by Jean-François Ouvrard and Camille Thubin (2020), “The structure of income helps to understand changes in the household saving ratio in France”, Bank of France Bulletin, n°227/9, January-February.

¹⁵ Bank of France (2019), Macroeconomic projections for France, box “Purchasing power gains initially allocated to savings and a gradual acceleration of consumption over time”, June

¹⁶ To cite only three references out of the abundant literature on the subject, see: Faure et alii (2012), already cited in the footnote 9; Céline Antonin, Mathieu Plane, Raul Sampognaro (2017), Was household consumer behaviour affected by the 2008 crisis? An econometric analysis of five major developed countries, OFCE Review, n°151; Ashoka Mody, Franziska Ohnsorge and Damiano Sandri (2012), Precautionary Savings in the Great Recession, IMF Working Paper n°12/42, February.

¹⁷ With a negative multiplier and lots of confidence, austerity can even be expansionist (theoretically, and with multiple conditions; empirically it is much harder to prove).

¹⁸ Antonin et alii (2017), previously cited in the footnote 16.

¹⁹ Mody et alii (2012), previously cited in the footnote 16; Karine Berger and Aurélien Daubaire (2003), Changes in the household savings rate in a few OECD countries: an interpretation based on medium-term determinants, Revue d’économie politique, vol. 113, November-December.

²⁰ Henri Fraisse (2004), Something new about the French household savings rate? Bulletin de la Banque de France, n°130, October; Bardaji et alii (2014), previously cited in the footnote 13.

²¹ Terms taken from Bardaji et alii (2014), op. cit..

²² Lucie Brice Mansencal, Patricia Croutte, and Sandra Hoibian, with the participation of Victor Prieur (2019), In forty years: more freedom, but also more worries, in France, social portrait, 2019 edition, INSEE Références.

²³ Tax increases in Q1 2018 (CSG, tobacco and fuel taxes) were stronger than tax cuts (transformation of ISF wealth tax into IFI, PFU flat tax, cut in employee healthcare and unemployment insurance contributions).

²⁴ Valérie Chauvin and Olivier Damette (2010), Wealth effects: the French case, Economie et Statistique, n°438-440; Antonin et alii (2017), op. cit..

²⁵ The savings rate of households in the 5th quintile is about 30%, compared to an average savings rate for all households of 16% in 2017.

²⁶ Maarten Dossche, Magnus Forsells, Luca Rossi and Grigor Stoevsky (2018), Private consumption and its drivers in the current economic expansion, ECB, Economic bulletin, n°5.