Financial effects in historic consumption and investment functions

ABSTRACT

The global financial crisis has highlighted the importance of financial factors on economic performance. Most of the existing research analyses the contemporary experience, and especially the 1980 s onwards. This paper investigates the effects of stock prices, real estate prices and debt on consumption and investment expenditures by estimating consumption and investment equations for about 110 years of data for Britain, France, Norway and Sweden. We find positive debt effects on consumption in three of four countries, but no consistent effects of stock prices and house prices. Effects on consumption are stronger in the more market-based economy of Britain than in the more state-oriented economies of France, Norway and Sweden: in Britain but not in the other countries, consumption has increased in response to stock and house prices. We find positive effects of stock prices on investment in all four countries. Credit expansion boosts investment in some of the countries while house prices seem to have become more influential for investment after 1980. The results indicate that there is interesting variation in financial effects over time, and point forward for possible research on the interaction between financial variables and the real economy since the early twentieth century.

KEYWORDS:

• Financial effects
• asset prices
• consumption
• investment
• macroeconomic history

JEL CLASSIFICATION:

• E21
• E22
• N13
• N14

1. Introduction

What causes variations in the central macroeconomic variables, consumption and investment? In a time of drastically growing house prices in the wealthy economies (Knoll, Schularick, and Steger 2017), as well as growing but fluctuating stock prices (Shiller 2005), the impact of asset prices on the real economy has become an important issue in economics. We may ask questions such as: what happens with consumption when house prices grow; or does growing household debt increase consumption, or do growing stock prices lead to higher investment rates? There is a lively literature on these issues. So Poterba and Samwick (1995, 295), writing during the economic boom in the US in the mid-1990 s, could state that, ‘The bull market of the last year has raised the total value of corporate stock in the United States by nearly a trillion dollars.’, and ask, among other questions, ‘How do rising stock prices affect consumer spending?’ In our times of great asset market growth and turbulence, this kind of question is important.¹

Most of the macroeconomic research on the effects of asset prices on consumption and investment, i.e. wealth effects, has used relatively short time series, focusing especially on the period since the 1970 s and 1980 s. There is a growing, but presently still modest research interest in earlier historic experience. Jorda et al (2016b) have presented data on debt and real estate prices for long time periods and Jordà, Schularick, and Taylor (2013) analyse the effect of credit growth on subsequent recessions. Stockhammer, Rabinovich, and Reddy (2018) estimate demand regimes using more than a hundred years of data, focussing on effects of income distribution and net private wealth. In this paper, as we use long macroeconomic time series from recent historical national accounts research to estimate consumption and investment equations with a century or more of data for four countries: Britain (1875–2015), France (1896–2016), Norway (1914–2016) and Sweden (1900–2016). The contribution of the paper is thus to present estimates of wealth and debt effects for much longer periods than the existing literature, covering very different macroeconomic regimes and contexts than the post-1980 deregulated, ‘financialized’ economies. The countries are chosen due to the availability of long-run data, but they also represent different types of growth regimes and financial systems. Britain is an archetypical case of a stock market-based financial system (Hall and Soskice 2001), while France is classified as a bank-based system.² Norway is placed in the market-based category while Sweden is in the bank-based category (Ludwig and Slok 2004). In terms of welfare state regimes, Britain is a liberal system, France a Continental European one, and Norway and Sweden are archetypical Social Democratic welfare regimes (Esping-Andersen 1990). Thus, the country sample, although relatively small, covers several distinct institutional configurations. Our combination of a rather diverse, if small, country sample and long time periods allows us to investigate wealth effects in a variety of settings.

The paper is structured as follows. Section 2 reviews the literature. Section 3 presents the data which we use. Section 4 presents our empirical results, and in section 5 we draw the conclusions.

2. Literature review: financial effects in historical perspective

2.1. Wealth effects on consumption

The standard approach to wealth effects is the life cycle model developed especially by Franco Modigliani in the 1960 s and 1970 s. Households are expected to vary their savings and consumption rates over the life cycle to keep their consumption and living standards relatively constant even as current income varies from the prime of working age to the pension income of a pensioner. In addition to current income, households also consume out of their wealth; the question is to which degree. Modigliani’s rule of thumb assumption was that households consume about five cents extra for one dollar more of wealth (Cf. Poterba 2000; Davis and Palumbo 2001). The five cents to the dollar rule of thumb still has support (Fair 2017), even though some researchers would lower the benchmark to three to four cents (Ludvigson and Steindel 1999).³

In the 2000 s, the literature on wealth effects has proliferated, focusing on two types of wealth: housing assets and stocks. Papers such as Poterba and Samwick (1995) were explicitly motivated by the post-1980 s stock market rallies, but interest in house prices has increased recently as housing has become a booming sector (cf. Grossman and Steger 2017). The then chairman of the Federal Reserve Alan Greenspan pointed out in 2001 that the marginal propensity to consume should be higher out of housing wealth than out of stock market wealth (Cf. Ludwig and Slok 2004.) A string of studies has since then considered both house prices and stock prices, with the possibility for differential effects of the two asset types.

Case, Quigley, and Shiller (2005) across countries and across US states since the 1980 s find significant effects of housing wealth on consumption (internationally 0.10, in the 0 USD.05) but not of stock prices. The opposite argument, of no housing wealth effects, is represented by Buiter (2010) who claims that ‘housing wealth isn’t wealth’. According to Buiter, the increased wealth of house-owners is counteracted by the relative loss of renters, so that increasing house prices in aggregate have no expansionary effect on the economy.⁴ Studies of effects of housing wealth do find that the interaction between house prices and borrowing matters, in the sense that households’ need to save for the down payment can restrain consumption (Lindner 2014), while borrowing against the rising price of the house can increase consumption (Mian and Sufi 2014). These studies tend to find larger effects of house prices after c. 1985, as financial deregulation has allowed for greater borrowing (Goodhart and Hofmann 2008; Slacalek 2009). There might also be differences between countries: especially, it has been suggested that consumption is more effected by stock prices and house prices in more market-based financial systems in the Anglo-Saxon countries. So Ludwig and Slok (2004) find higher responsiveness of consumption changes in stock prices in countries with a market-based financial system than in countries with bank-based systems (also Slacalek 2009), and Aron et al. (2012) find larger effects of credit on consumption in Britain and the US than in Japan.

Like asset prices, the interest in household debt among researchers has increased since the 1990 s, as household indebtedness has grown (cf. Zinman 2015). The debt stock is expected to restrain consumption by increasing interest payments, which decreases the incomes disposable for new consumption. At the same time, the flow of debt can increase consumption, as people take up new loans to consume. In other words, it is essential to investigate this variable’s effect on consumption and investment in levels as well as in differences. Several studies have emphasized household debt and how, when house prices fall, the deleveraging of households deepen recessions (Eggertsson and Krugman 2012; Mian, Rao, and Sufi 2013; Jordà, Schularick, and Taylor 2016a). Household debt is intimately related to asset prices, especially house prices, and Slacalek (2009) argues that house prices have become more important determinants of investment since the 1980 s precisely because of the growing possibility to mortgage one’s house.

The extant literature on wealth effects almost exclusively focuses on the present period, going back to the 1980 s or so. However, there are good reasons to expect that wealth effects may vary over time. Poterba and Samwick (1995, 341) claim that the marginal propensity to consume out of wealth was low in the interwar period. One key reason why effects may vary over time is the changing patterns in asset ownership. One researcher has claimed that by 1870, Britain had become a ‘stock-and-bondholding aristocracy, measuring income in dividends and wealth in the quotations of the Stock Exchange’ (cited in Rutterford et al. 2011, p. 127.). We may on the other hand question how broadly shares are held and how the consumption functions of the share-holding groups differ from the average in the population. Acheson, Campbell, and Turner (2017) map the rather exclusive group who held shares in Victorian Britain. Poterba and Samwick (1995) show that stock ownership in the US from 1983 to 1992 was still very unequal, but decreasing. The 0.5 per cent of stockholders who owned the most stock owned 55 percent of the total in 1983, and 37 percent in 1992. In other words, we may expect stronger effects of stock price appreciation on consumption in the late twentieth century than in the first half of the century, as the ownership of stocks was likely more widespread in the latter period. On the other hand, we may also expect weaker effects of stock prices than of house prices even in the latter period, as stock ownership is more unequal than home ownership.

2.2. Investment

There are several reasons to expect effects of asset prices, especially stock prices, on investment. Morck et al. (1990) discuss four hypotheses.⁵ One, the passive informant hypothesis says that the stock market reflects information about the investment environment, without in itself adding new knowledge. For this reason, it will be correlated with investment. Second, the active informant hypothesis says that managers ‘rely on the stock market as a source of information when making investment decisions’ (Gjerde, Knivsflå, and Sættem 2001, 566). Three, the stock market can affect investment by affecting the cost of new stock issuance. Four, the stock market pressure hypothesis says that the stock market evaluation affects managers’ behaviour by affecting their incentives. In Chapter 12 of the General Theory, Keynes used precisely the value of share prices on the stock exchange as the measure of expected profitability of investment (Lopez and Mott 1999, 297).

It has been debated in economics to which degree profitability affects corporate investment. In Hall and Jorgenson (1967) neoclassical model, the relevant variable of consideration for corporations was a cost of capital which was set in the securities markets, i.e. external to the firm (Fazzari et al. 1988, p. 144–145; Gordon 1992). This economy-wide cost of capital will be measured in our empirical investigation by the real interest rate.

The empirical evidence is not very clear cut. On the interest rate, neither Kothari, Lewellen, and Warner (2016) nor Gordon and Veitch (1986) find any effect on investment in the US. Blanchard, Rhee, and Summers (1993) investigate the relationship between stock market valuation and investment from the 1920 s to the 1980 s in the United States, and find that there is very little relation between the two; the same non-result for profits is found in a study of Norway 1952–1995 (Gjerde, Knivsflå, and Sættem 2001) while a study of Dutch industrial investment patterns 1950–1987 found that profits indeed lead investments (Van Ees et al 1997). Kothari, Lewellen, and Warner (2016) investigate corporate investment in the 1952 USD–2010 and find that higher stock prices are associated with more investment but Gordon and Veitch (1986) investigate investment in the US business cycle 1919–1983 and claim that investment, especially buildings investment, in line with a Keynesian view, is very hard to pin down econometrically. They argue that investment is an autonomous factor and drives the business cycle. The investment boom of the 1920 s and subsequent slump of the 1930 s was in their opinion influenced by speculation and ‘animal spirits’ a la Keynes, as well as a ‘Schumpeterian’ factor of a bundling of innovations which broke through in that decade (Gordon and Veitch, pp. 326–327).

2.3. The country context

As shown in several studies, wealth effects might differ over time. The same goes for the country context: we might assume that in countries with a larger place role for the market and less for the state in coordination, wealth effects will be more marked. In our study, this will especially be a case of Britain vis-à-vis the more state-oriented economies of France, Norway and Sweden. Peter Hall emphasized that ‘since 1918, industries in Britain have essentially been asked to rationalize themselves’, while in France, the state in the postwar period took a very active role in steering industry, with nationalized banks and the Bank of France actively steering private investment (Hall 1986, pp. 51, 153). Equally, in Sweden in the postwar period, the finance ministry and the politically steered central bank did not hesitate to approach a semi-planned economy, with very strong regulation of profits and investments (Pontusson 1992); Katzenstein (1985, 61) in a classic study refers to this as ‘indicative planning’ while noting that in Norway, interventionism was even more advanced. In both France and Sweden, and especially in Sweden (Belfrage and Kallifatides 2018), such regulations and planning institutions have been dismantled since the early 1980 s. Thus, we have good reasons to expect different financial effects between countries and across time, especially on investment.

3. Data and method

In the wealth effects literature, two types of wealth measures are used. One is the direct ownership of households, derived from household surveys (i.e. Poterba and Samwick 1995; Dynan and Maki 2001). The other, utilized here, is asset price indices, as used by for example Ludwig and Slok (2004). This is of course not a direct measure of wealth, but the growth of the house price index and the stock price index should capture the growth of these two types of wealth well, because prices will usually are more volatile than quantities (cf. Lettau and Ludvigson 2004).

Beyond the two types of wealth as our major independent variables, we also use control variables. GDP is used as the indicator of income.⁶ The real interest rate is a key macroeconomic variable which is usually controlled for in the literature, even if its estimated effects typically are small or non-existent (Campbell and Mankiw 1991; Slacalek 2009).

We utilize data compiled in the last ten to fifteen years by economic historians working within extensive projects on historical national accounts, sponsored not the least by central banks. A very important sources is also the Jordà, Schularick, and Taylor (2016b) Macrohistory database which has very rich information on credit and on house prices. To update all series to 2016, we have linked with other sources, especially from statistical agencies. See Appendix for a detailed explanation.

Our baseline debt variable in the consumption models is mortgage debt, while in the investment models it is business debt. We also test use total debt to ascertain that the results do not change.

3.1. Econometric methodology

Our baseline consumption and investment equations are

$C=f\left(Y,PH,PS,DHH\right)$

With expected signs C_Y, C_PH, C_PS > 0, C_DHH < 0

$I=g\left(Y,PH,PS,DBUS,i\right)$

With expected signs I_Y, I_PH, I_PS > 0, I_DBUS < 0

Where C, I, Y, PH, PS, DHH, DBUS and i stand for consumption, investment, GDP, house prices, stock prices, household debt, business debt and the interest rate respectively. All variables are inflation adjusted.

The macroeconomic variables that we work with are almost all trended, which is also revealed by a standard test for unit roots, the augmented Dickey-Fuller Test. For this reason we must account for the stochastic trend. We use as error correction models where we find evidence that the variables are cointegrated (cf. Banerjee, Dolado, and Mestre 1998)⁷ and first difference specifications otherwise. We experiment with the lag structure and allow for up to two lags of the differences. All variables involved bar the interest rate are in logarithm form. The full error correction model for consumption is:

(1) $\begin{array}{rl}& \mathrm{\Delta }\mathrm{l}\mathrm{o}\mathrm{g}\left(C\right)=\rho \mathrm{l}\mathrm{o}\mathrm{g}(C_{t-1})+{\beta }_{1}log(Y_{t-1})+{\beta }_2\mathrm{l}\mathrm{o}\mathrm{g}\left(DH{H}_{t-1}\right)+{\beta }_{3}log(P{S}_{t-1})+{\beta }_{4}log(P{H}_{t-1})\\ & +{\sigma }_{1n}{\sum }_n\mathrm{\Delta }log\left(Y_{t-n}\right)+{\sigma }_{2n}{\sum }_n\mathrm{\Delta }log\left(DH{H}_{t-n}\right){\sigma }_{3n}{\sum }_n\mathrm{\Delta }log\left(P{S}_{t-n}\right)\\ & +{\sigma }_{4n}{\sum }_n\mathrm{\Delta }log\left(P{H}_{t-n}\right)+\alpha +e_t\end{array}$(1)

For investment, we use the same model but with the real interest rate included, and with corporate debt instead of household debt. The long-run effects are calculated as the coefficient divided by the coefficient of the lagged dependent variable, β_j/-ρ in the terms of Equation 1 (De Boef and Keele 2008). When the variables are not cointegrated, we use a distributed lag version of first difference models. We have experiment with different lag structures with and without contemporaneous effects and up to three lags of all independent variables. Thereafter, we have tested down the models to simplify specification. Overall, effects seem to quickly wither away after one year’s lag.⁸ We have decreased the number of lags to the ones which are statistically significant at least on the 10 per cent level.

Year dummies are used for years during the World Wars when the macroeconomic variables were strongly influence by wartime measures. The year dummies have been chosen based on a first stage estimation for the full sample with dummies for each of the war years and the immediate aftermath (1914–1920 and 1939–1946). Those dummies which were statistical significanct at the 10 per cent level or higher are retained.

For each country we present several specifications: a baseline model (tested down) without any period interactions, then interactions of the financial variables with dummy variables for the postwar (1945–80) and post-1980 (1980–2016) dummies, one at a time, and then a model with all financial variables (stock prices, house prices, debt) interacted with the period dummies. We will refer to these interactions as period-specific effects.

4. Results

Table 1 show the results for consumption for Britain. We fail to find evidence for cointegration, so we estimate first differences models. Specifications 2 through 5 include contemporaneous effects and period-specific effects for each of the three key independent variables; first one at a time (specifications 2–4) and then interactions for all three in the same specification (5). Specification 6 drops the contemporaneous effects as these might suffer from endogeneity problems; this does not change the substantive results. Results are very similar across specifications. We thus prefer specification (1) as more parsimonious. We consistently find statistically significant effects for share prices and for house prices. For example in specification (1) share prices (with one lag) are statistically significant at the 1% level and house prices are statistically significant at the 5% level contemporaneously. The respective coefficient estimates for stock prices are 0.04 and for house prices 0.05 and 0.04, giving a total effect of 5 per cent for stock prices and 9 per cent for house prices. This means that coefficient estimates are in similar order of magnitude as previous research (for more recent time periods) has reported. We also find statistically significant effects of household debt. These effects are positive for contemporaneous effects and negative for the first lag. The positive effects are somewhat larger, so that the total effect is around +0.06. This suggests that households are credit constrained.

Table 1. Wealth effects on consumption in Britain.

	(1)	(2)	(3)	(4)	(5)	(6)
Δ Y t	0.37***	0.30***	0.26***	0.35***	0.27***
	(4.93)	(3.88)	(3.59)	(4.75)	(3.78)
Δ Y t − 1	−0.17**	−0.01	−0.01	−0.18**	−0.13*	0.02
	(−2.40)	(−0.25)	(−0.16)	(−2.58)	(−1.82)	(0.29)
Δ Y t-2	0.12***			0.13***	0.11**	0.05
	(2.86)			(2.97)	(2.61)	(0.98)
Δ Stock prices t-1	0.04***	0.05***	0.04***	0.04***	0.06***	0.05***
	(5.06)	(4.68)	(4.98)	(4.78)	(5.69)	(5.03)
Δ Stock prices t-1 * postwar		−0.03			−0.04**
		(−1.33)			(−2.19)
Δ Stock prices t-1 * post-1980		−0.02			−0.04*
		(−1.18)			(−1.85)
Δ House prices t	0.05**	0.05**	−0.04	0.04*	−0.04
	(2.45)	(2.35)	(−1.25)	(1.90)	(−1.19)
Δ House prices t * postwar			0.14***		0.12**
			(2.80)		(2.18)
Δ House prices t * post-1980			0.18***		0.19***
			(3.74)		(3.73)
Δ House prices t-1	0.04**	0.04*	0.01	0.03	0.01	0.07***
	(2.00)	(1.67)	(0.27)	(1.39)	(0.64)	(2.84)
Δ Household debt t	0.13***	0.14***	0.12***	0.11***	0.15***
	(3.80)	(3.96)	(3.87)	(2.93)	(4.01)
Δ Household debt t * postwar				0.10	0.07
				(1.05)	(0.71)
Δ Household debt t * post-1980				0.03	−0.03
				(0.67)	(−0.55)
Δ Household debt t-1	−0.07**	−0.10***	−0.09***	−0.06*	−0.09***	0.02
	(−2.31)	(−3.12)	(−3.01)	(−1.93)	(−2.92)	(0.89)
Postwar dummy		0.01**	0.01**	0.00	0.00
		(2.18)	(2.30)	(0.41)	(0.75)
Post-1980 dummy		0.01**	0.01	0.01	0.01*
		(2.12)	(1.63)	(1.42)	(1.77)
Constant	0.01***	0.01**	0.01***	0.01*	0.01**	0.02***
	(4.02)	(2.47)	(2.70)	(1.86)	(2.15)	(6.60)
AIC	−638	−633	−647	−640	−653	−591
Durbin-Watson d	1.87	2.04	2.07	1.95	1.91	1.83
Observations	117	117	117	117	117	118
Adjusted R-squared	0.79	0.79	0.81	0.80	0.83	0.66

t-statistics in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. Year dummies for 1915, 1916, 1917 1919, 1920, 1941, 1945 and 1946 not reported.

Table 2 summarises results for investment. Again we fail to find sufficient evidence for cointegration and use first differences models. Because of autocorrelation problems we include a lagged dependent variable. We find strong contemporaneous demand effects. The expected positive effect of stock prices on investment are only found in three of the five specifications; the coefficient varies between 0.07 and 0.10. House prices have statistically significant negative (i.e. perverse) effects which are quite large, around −0.33. Credit expansion only has a statistically significant effect in one of the specifications. Interest rates have a small, but statistically significant negative effect on investment. The period-specific effects are limited to that stock prices have less effect after 1980 (specifications 2, 5) and that house prices appear to have a positive effect after 1980, unlike before that (specifications 3, 5). We conclude that overall, wealth effects on investment are weak in Britain. Given the changes in British political economy from the postwar period to the post-Thatcher period (Hall 1986), it is surprising that the effects do not vary more over time.

Table 2. Wealth effects on investment in Britain.

	(1)	(2)	(3)	(4)	(5)
Δ Investment t − 1	0.34***	0.32***	0.28***	0.29***	0.28***
	(5.66)	(4.93)	(4.43)	(4.41)	(4.40)
Δ Y t	1.08***	1.05***	0.84***	0.92***	0.85***
	(3.62)	(3.51)	(2.82)	(2.98)	(2.75)
Δ Y t − 1	0.27	0.29	0.06	0.19	0.11
	(0.89)	(0.94)	(0.18)	(0.62)	(0.36)
Δ Y t-2	−0.55***	−0.57***	−0.50***	−0.55***	−0.54***
	(−3.25)	(−3.36)	(−3.02)	(−3.22)	(−3.13)
Δ Interest rate t − 1	−0.01***	−0.01***	−0.01***	−0.01***	−0.01***
	(−4.68)	(−4.55)	(−4.84)	(−4.49)	(−4.74)
Δ Interest rate t-2	−0.00***	−0.00***	−0.00***	−0.00***	−0.00***
	(−2.72)	(−2.76)	(−3.32)	(−3.10)	(−3.24)
Δ Stock prices t-2	0.05	0.09**	0.07**	0.05	0.10**
	(1.46)	(2.23)	(2.02)	(1.41)	(2.45)
Δ Stock prices t-2 * postwar		−0.05			−0.04
		(−0.64)			(−0.45)
Δ Stock prices t-2 * post-1980		−0.16**			−0.16*
		(−1.99)			(−1.90)
Δ House prices t	0.08	0.06	0.04	0.09	−0.01
	(0.77)	(0.55)	(0.25)	(0.90)	(−0.07)
Δ House prices t * postwar			−0.29		−0.20
			(−1.33)		(−0.79)
Δ House prices t * post-1980			0.37*		0.36
			(1.75)		(1.47)
Δ House prices t-1	−0.32***	−0.31***	−0.33***	−0.35***	−0.34***
	(−3.25)	(−3.12)	(−3.35)	(−3.46)	(−3.34)
Δ Business debt t	0.10	0.10	0.14*	0.16	0.16
	(1.27)	(1.24)	(1.75)	(1.18)	(1.12)
Δ Business debt t * postwar				−0.18	−0.05
				(−1.06)	(−0.27)
Δ Business debt t * post-1980				0.10	0.05
				(0.59)	(0.24)
Postwar dummy		0.02	0.03**	0.03*	0.03**
		(1.39)	(2.20)	(1.98)	(2.06)
Post-1980 dummy		0.01	−0.01	0.00	−0.00
		(0.67)	(−0.42)	(0.00)	(−0.20)
Constant	−0.00	−0.01	−0.00	−0.00	−0.00
	(−0.34)	(−0.78)	(−0.11)	(−0.45)	(−0.12)
AIC	−304	−304	−311	−303	−308
Durbin-Watson d	1.94	1.91	1.97	1.99	1.99
Observations	117	117	117	117	117
Adjusted R-squared	0.78	0.78	0.80	0.78	0.80

t-statistics in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. Year dummies for 1915, 1916, 1917, 1919, 1920, 1940. Including debt-to-GDP ratio does not make a difference; see Appendix Table Y.

For France (Table 3) we estimate consumption first as function of total private debt, and then in specifications 6–7 use only household debt.⁹ Specification 1 is the baseline; specifications 2–4 interacts the financial variables with the period dummies one at a time, and specification 5 includes interactions for all three. The t-value of the error correction term is high, and the error correction term is in the plausible range. The long run effects of stock prices differ over time. Overall the effect is estimated as negative (around −0.06), but when period interactions are included in models 2, 5 and 7, we see that the estimated effect after 1945 is positive (0.05 to 0.09), and this is also true after 1980, when the effect is stronger (0.10–0.14). In the short run, stock price growth stimulate consumption by about 0.13.¹⁰ House prices stimulate consumption according to four of our seven specification; this is especially true after 1945. The estimated effects are about as large as those for stock prices. The short-run effects of house prices are contradictory. The stock of outstanding private debt has no consistently significant effect. Effects of debt growth are unstable – only in the two first specifications are there significant positive effects, at 0.06. Surprisingly, household debt specifically (specifications 6 and 7) have no effect on consumption at all.

France is typically seen as a very different type of capitalism to Britain – less stock market driven, and less financialized. However, here we find little evidence for such assumptions of the French economy as less stock market-driven than the British (Hall and Soskice 2001). The estimated effects of stock prices are about the same in Britain and France, at least after 1945 – before that, stock prices mattered for consumption in Britain but not in France.

Table 4 reports the investment results for France. There is evidence for cointegration; thus we use error correction models. As with consumption, specification 1 is the baseline; specifications 2–4 interacts the financial variables with the period dummies one at a time, and specification 5 includes interactions for all three. Stock price growth increases investment in the short run, by about 0.15. Debt expansion also expands investment, by about 0.28; when we look only at business debt, it is still significant, but around 0.18. The smaller effect indicates that mortgage credit also has a strongly expansive effect on investment (presumably housing-related). There are no stable long run effects.

Table 3. Wealth effects on consumption in France.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Consumption t-1	−0.36***	−0.58***	−0.59***	−0.74***	−0.80***	−0.36***	−0.52***
	(−3.71)	(−4.12)	(−3.68)	(−4.19)	(−4.65)	(−3.33)	(−3.73)
Y t-1	0.32***	0.34**	0.49***	0.59***	0.51***	0.32***	0.37***
	(2.95)	(2.63)	(3.32)	(3.81)	(3.33)	(2.95)	(2.90)
Stock prices t-1	−0.00	−0.04***	−0.01	−0.02*	−0.05***	−0.01	−0.04**
	(−0.59)	(−3.17)	(−0.81)	(−1.79)	(−3.55)	(−0.64)	(−2.47)
Stock prices t-1 * postwar		0.05**			0.04*
		(2.28)			(1.94)
Stock prices t-1 * post-1980		0.08***			0.08***		0.05*
		(3.44)			(3.06)		(1.85)
House prices t-1	0.01	0.05***	−0.00	0.01	0.05**	0.01	0.07**
	(0.86)	(3.11)	(−0.14)	(0.75)	(2.56)	(0.77)	(2.55)
House prices t-1 * postwar			0.03*		−0.03
			(1.91)		(−0.83)
House prices t-1 * post-1980			0.05*		−0.04		−0.07
			(1.67)		(−0.78)		(−0.92)
Debt t-1	−0.01	0.02	−0.00	−0.00	0.02
	(−0.51)	(0.82)	(−0.12)	(−0.19)	(0.72)
Debt t-1 * postwar				0.04**	0.07
				(2.42)	(1.63)
Debt t-1 * post-1980				0.09**	0.08
				(2.46)	(1.30)
Household debt t-1						−0.01	−0.02*
						(−0.79)	(−1.81)
Household debt t-1 * post-1980							0.07
							(1.27)
Δ Y t-1	−0.25**	−0.19*	−0.25**	−0.23**	−0.19*	−0.05	0.05
	(−2.25)	(−1.72)	(−2.17)	(−2.04)	(−1.73)	(−0.33)	(0.34)
Δ Stock prices t-1	0.01	0.01	0.01	0.02	0.01	0.01	0.02
	(0.48)	(0.44)	(0.69)	(0.95)	(0.59)	(0.37)	(0.76)
Δ House prices t-1	0.10*	0.04	0.08	0.09*	0.05	0.16**	0.16**
	(1.96)	(0.70)	(1.63)	(1.90)	(0.99)	(2.18)	(2.11)
Δ House prices t-2	−0.12***	−0.16***	−0.11**	−0.09**	−0.12***	−0.16***	−0.16***
	(−2.82)	(−3.62)	(−2.49)	(−2.12)	(−2.70)	(−2.85)	(−2.82)
Δ Debt t-1	0.06*	0.06*	0.05	0.05	0.05
	(1.73)	(1.73)	(1.62)	(1.48)	(1.57)
Δ Household debt t-1						−0.07	−0.05
						(−1.08)	(−0.80)
Postwar dummy		−0.23**	−0.11*	−0.34**	−0.66**
		(−2.36)	(−1.95)	(−2.48)	(−2.57)
Post-1980 dummy		−0.30***	−0.18	−0.70**	−0.80**		−0.41**
		(−3.22)	(−1.46)	(−2.37)	(−2.22)		(−2.39)
Constant	0.17*	1.08***	0.46***	0.73***	1.40***	0.14	0.80***
	(1.86)	(3.99)	(2.71)	(3.31)	(4.65)	(1.24)	(2.93)
AIC	−408	−416	−407	−411	−416	−354	−354
Durbin-Watson d	2.16	2.11	2.07	2.05	2.13	2.14	2.10
Observations	98	98	98	98	98	99	99
Adjusted R-squared	0.61	0.65	0.62	0.63	0.66	0.20	0.24
t-statistics in parentheses

***p < 0.01, **p < 0.05, *p < 0.1

Table 4. Wealth effects on investment in France.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Investment t-1	−0.33***	−0.37***	−0.40***	−0.40***	−0.47***	−0.48***	−0.67***
	(−3.57)	(−3.86)	(−4.16)	(−4.12)	(−4.15)	(−4.39)	(−5.75)
Y t-1	0.07	0.29	0.22	0.25*	0.43**	0.20**	0.55***
	(0.70)	(1.59)	(1.64)	(1.68)	(2.01)	(2.20)	(3.19)
Interest rate t-1	−0.00	−0.00	−0.00	−0.00	−0.00	−0.00*	−0.00*
	(−0.86)	(−0.54)	(−0.81)	(−0.78)	(−0.30)	(−1.96)	(−1.76)
Stock prices t-1	−0.01	0.00	−0.00	0.01	0.05	−0.03*	−0.01
	(−0.75)	(0.03)	(−0.07)	(0.21)	(1.00)	(−1.72)	(−0.13)
Stock prices t-1 * postwar		0.04			−0.03		0.03
		(0.72)			(−0.41)		(0.45)
Stock prices t-1 * post-1980		−0.02			−0.05		0.00
		(−0.35)			(−0.68)		(0.01)
House prices t-1	0.04	0.02	0.01	0.01	−0.06	0.07**	0.01
	(1.35)	(0.48)	(0.16)	(0.28)	(−0.92)	(2.07)	(0.15)
House prices t-1 * postwar			0.05		0.18		0.12
			(1.32)		(1.57)		(1.14)
House prices t-1 * post-1980			0.00		0.19		0.15
			(0.01)		(1.14)		(1.37)
Debt t-1	0.14	0.07	0.12	0.11	0.12
	(1.65)	(0.78)	(1.38)	(1.27)	(1.26)
Debt t-1 * postwar				0.04	−0.17
				(0.81)	(−1.23)
Debt t-1 * post-1980				−0.04	−0.21
				(−0.50)	(−1.06)
Business debt t-1						0.15***	0.12**
						(2.68)	(2.10)
Business debt t-1 * postwar							−0.08
							(−0.62)
Business debt t-1 * post-1980							−0.18
							(−1.08)
Δ Y t − 1	0.37	0.19	0.18	0.19	0.13	0.62*	0.38
	(1.20)	(0.58)	(0.57)	(0.60)	(0.41)	(1.97)	(1.31)
Δ Y t − 2	0.79**	0.67**	0.65**	0.65**	0.60*	0.62*	0.85***
	(2.57)	(2.13)	(2.10)	(2.08)	(1.88)	(1.95)	(2.87)
Δ Interest rate t-1	0.00	0.00	0.00	0.00	0.00	0.01**	0.00**
	(1.36)	(1.17)	(1.35)	(1.29)	(0.84)	(2.57)	(2.49)
Δ Stock prices t − 1	0.14**	0.15**	0.15**	0.14**	0.15**	0.12**	0.13**
	(2.54)	(2.58)	(2.60)	(2.29)	(2.37)	(2.06)	(2.33)
Δ House prices t − 1	0.11	0.02	0.08	0.07	0.03	−0.00	0.01
	(1.01)	(0.20)	(0.68)	(0.61)	(0.21)	(−0.01)	(0.09)
Δ Debt t − 1	0.27***	0.29***	0.27***	0.28***	0.28***
	(3.16)	(3.32)	(3.14)	(3.19)	(3.12)
Δ Business debt t − 1						0.17**	0.18***
						(2.36)	(2.78)
Postwar dummy		−0.15	−0.13	−0.21	0.75		0.07
		(−0.54)	(−0.93)	(−0.65)	(0.96)		(0.11)
Post-1980 dummy		0.04	−0.04	0.33	1.08		0.68
		(0.15)	(−0.11)	(0.45)	(1.01)		(0.70)
Constant	0.02	−0.71	−0.44	−0.61	−1.39	−0.05	−1.10
	(0.07)	(−0.99)	(−1.05)	(−1.18)	(−1.59)	(−0.20)	(−1.51)
AIC
Durbin-Watson d
Observations	97	97	97	97	97	98	92
Adjusted R-squared	0.44	0.45	0.46	0.45	0.45	0.33	0.54
t-statistics in parentheses

***p < 0.01, **p < 0.05, *p < 0.1

None of the effects of stock prices, house prices or debt vary over time – there are no such results stable across specifications. This is different from the results for Britain and indicates a greater stability in macroeconomic regime from the postwar years to the post-1980 period. Of course, French capitalism is not today the dirigiste regime of the 1950 s and 1960 s (Amable, Guillaud, and Palombarini 2012), but neither has it experienced a regime change as drastic as Thatcherism. Another interesting difference between Britain and France is that both consumption and investment behave more idiosyncratically in the latter country: the share of variation that our models can explain is much higher in Britain. (This despite the fact that we use first difference models for Britain and ECMs for France.) In Britain the adjusted R-squared for consumption and investment is around 0.80 while in France for consumption it’s around 0.65 and for investment only around 0.50 (after 1958, when the sample is constrained by our use of the business debt variable, it is only 0.20). There is more unexplained variation in France, possibly due to the larger importance of regulation and state interventionism there (cf. Hall 1986).

Table 5 reports consumption results for Norway. We fail to find evidence for cointegration and apply difference specifications. Income is the most important driver and the contemporaneous effect is very large, around 0.80. Stock price growth increases consumption by about 0.02 and house price growth by 0.10. Household debt has no positive significant effect. In one specification (4), in the postwar era, the effect of credit growth on consumption was negative; however, this does not hold in another specification (5). In fact, the only time-varying effect which holds across specifications is that stock prices have no positive effect on consumption after 1980. Landsem (2016) in a study of Norway since 1975 found that a one percentage point increase in housing wealth increased consumption by 0.08 percent, while the same size increase in financial wealth in the shape of stocks and bonds increased consumption by 0.27 percent. Brodin and Nymoen (1992) on the other hand, with a quarterly sample 1968–1989 found that household wealth increased consumption by 0.27 p.p in the long run. We ony examine the short run, but it does seem, as in Landsem’s study, that house prices have more effect on consumption in Norway than stock prices do.

Table 5. Wealth effects on consumption in Norway.

	(1)	(2)	(3)	(4)	(5)
Δ Y t	0.80***	0.79***	0.77***	0.79***	0.78***
	(11.73)	(10.75)	(10.05)	(10.67)	(10.21)
Δ Stock prices t	0.02**	0.02*	0.02*	0.02*	0.02*
	(2.05)	(1.70)	(1.90)	(1.96)	(1.70)
Δ Stock prices t-1	0.00	0.03*	0.00	0.00	0.03*
	(0.22)	(1.77)	(0.38)	(0.36)	(1.86)
Δ Stock prices t-1 * postwar		−0.05			−0.03
		(−1.28)			(−0.70)
Δ Stock prices t-1 * post-1980		−0.05**			−0.05**
		(−2.10)			(−2.09)
Δ Stock prices t-2	0.03***	0.03**	0.03**	0.03**	0.03**
	(2.69)	(2.31)	(2.55)	(2.55)	(2.23)
Δ House prices t	0.10***	0.10***	0.17***	0.09***	0.15**
	(3.70)	(3.81)	(3.19)	(3.05)	(2.61)
Δ House prices t * postwar			−0.07		−0.07
			(−0.99)		(−0.92)
Δ House prices t * post-1980			−0.11*		−0.09
			(−1.69)		(−1.28)
Δ Household debt t	0.01	0.01	−0.02	0.01	0.00
	(0.20)	(0.36)	(−0.43)	(0.21)	(0.06)
Δ Household debt t-1	0.00	0.00	−0.00	0.03	0.01
	(0.09)	(0.12)	(−0.13)	(0.74)	(0.19)
Δ Household debt t-1 * postwar				−0.15*	−0.12
				(−1.88)	(−1.41)
Δ Household debt t-1 * post-1980				−0.03	0.03
				(−0.43)	(0.40)
Postwar dummy		0.00	0.01	0.01*	0.01
		(0.38)	(1.03)	(1.89)	(1.49)
Post-1980 dummy		0.00	0.00	0.00	0.00
		(0.64)	(0.80)	(0.62)	(0.47)
Constant	−0.00	−0.00	−0.00	−0.00	−0.00
	(−0.66)	(−0.70)	(−0.48)	(−1.14)	(−0.54)
AIC	−499	−498	−496	−497	−496
Durbin-Watson d	1.84	1.91	1.79	1.79	1.77
Observations	100	100	100	100	100
Adjusted R-squared	0.81	0.81	0.81	0.81	0.81

t-statistics in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. Year dummies for 1917, 1919, 1940, 1944, 1945, 1946 not reported.

Results for investment are reported in Table 6. Again we apply difference specifications because of lack of evidence for cointegration. There is a very strong accelerator effect, with contemporaneous effects of income above 1. Stock prices increase investment by about 0.25, spread out over immediate effects and the one year lag; the effect of the one year lag is statistically significant. Neither house prices nor business debt have statistically significant effects. Gjerde, Knivsflå, and Sættem (2001) found no effects of the stock market on investments in Norway, in a study of 1952–1995. We on the other hand find an effect around 0.25, much in line with France and with Britain after 1980. As in Britain, the effect is lesser after 1980, when stock markets have boomed without a corresponding investment boom.

Table 6. Wealth effects on investment in Norway.

	(1)	(2)	(3)	(4)	(5)
Δ Y t	1.85***	1.83***	1.82***	1.77***	1.92***
	(7.71)	(7.04)	(6.39)	(6.72)	(6.71)
Δ Y t-1	0.34	0.24	0.26	0.26	0.25
	(1.26)	(0.86)	(0.90)	(0.89)	(0.89)
Δ Interest rate t	−0.01	−0.00	−0.01	−0.01	−0.01
	(−0.97)	(−0.44)	(−1.34)	(−1.07)	(−0.66)
Δ Interest rate t-1	−0.01	−0.01	−0.00	−0.00	−0.00
	(−0.92)	(−1.07)	(−0.34)	(−0.46)	(−0.23)
Δ Stock prices t	0.06	0.06	0.07	0.07	0.07
	(1.24)	(1.37)	(1.40)	(1.45)	(1.50)
Δ Stock prices t-1	0.16***	0.25***	0.17***	0.19***	0.25***
	(3.63)	(3.39)	(3.48)	(3.83)	(3.35)
Δ Stock prices t-1 * postwar		0.07			0.11
		(0.42)			(0.70)
Δ Stock prices t-1 * post-1980		−0.17*			−0.18*
		(−1.78)			(−1.79)
Δ House prices t	0.12	0.14	0.01	0.14	−0.03
	(1.21)	(1.37)	(0.05)	(1.31)	(−0.15)
Δ House prices t * postwar			0.02		−0.02
			(0.07)		(−0.07)
Δ House prices t * post-1980			0.32		0.41
			(1.31)		(1.62)
Δ House prices t-1	0.07	0.06	0.04	0.07	−0.01
	(0.68)	(0.64)	(0.39)	(0.67)	(−0.12)
Business debt t-1	−0.08	−0.05	−0.06	0.04	0.13
	(−0.89)	(−0.55)	(−0.67)	(0.30)	(0.95)
Business debt t-1 * postwar				−0.16	−0.26
				(−0.80)	(−1.33)
Business debt t-1 * post-1980				−0.24	−0.27
				(−1.13)	(−1.30)
Business debt t-2	0.06	0.04	0.03	0.04	−0.01
	(0.65)	(0.40)	(0.38)	(0.37)	(−0.15)
Postwar dummy		0.01	0.01	0.01	0.02
		(0.32)	(0.39)	(0.48)	(0.69)
Post-1980 dummy		−0.01	−0.02	−0.01	−0.01
		(−0.29)	(−0.88)	(−0.58)	(−0.52)
Constant	−0.05***	−0.04**	−0.04**	−0.04**	−0.04**
	(−3.75)	(−2.16)	(−2.35)	(−2.12)	(−2.15)
AIC	−227	−226	−223	−222	−225
Durbin-Watson d	2.30	2.41	2.35	2.33	2.50
Observations	101	101	101	101	101
Adjusted R-squared	0.84	0.85	0.84	0.84	0.85

t-statistics in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. Year dummies for 1918, 1919, 1920, 1944, 1945, 1946 not reported.

Sweden is a small open economy much like the Norwegian, but with a higher degree of financialization and more stock market based (Ludwig and Slok 2004). Table 7 shows the results for consumption. Again, we do not find cointegration, so use first differences specifications. While we find statistically significant effects for income, other level variables seem to have little effects. We find small negative, i.e.perverse, effects of stock prices on consumption.

Table 7. Wealth effects on consumption in Sweden.

	(1)	(2)	(3)	(4)	(5)
Δ Y t	0.98***	0.95***	0.96***	0.97***	0.95***
	(11.55)	(11.13)	(10.55)	(10.49)	(11.01)
Δ Stock prices t	−0.03***	−0.03***	−0.03**	−0.03**	−0.03***
	(−2.78)	(−3.08)	(−2.31)	(−2.34)	(−3.04)
Δ Stock prices t-1	−0.01	0.03	−0.01	−0.01	0.03
	(−1.00)	(1.20)	(−0.85)	(−0.78)	(1.26)
Δ Stock prices t-1 * postwar		0.03			0.03
		(1.06)			(0.95)
Δ Stock prices t-1 * post-1980		−0.07***			−0.08***
		(−3.02)			(−3.19)
Δ Stock prices t-2	−0.01	−0.01	−0.00	−0.00	−0.01
	(−0.57)	(−0.73)	(−0.30)	(−0.35)	(−0.56)
Δ House prices t	0.00	0.02	0.01	0.00	0.02
	(0.01)	(0.76)	(0.32)	(0.08)	(0.63)
Δ House prices t-1	0.02	−0.00	0.00	0.02	0.01
	(0.51)	(−0.04)	(0.02)	(0.61)	(0.12)
Δ House prices t-1 * postwar			0.07		0.01
			(0.91)		(0.08)
Δ House prices t-1 * post-1980			0.02		−0.06
			(0.27)		(−0.76)
Δ Household debt t-1	0.00	0.00	0.00	0.00	−0.00
	(0.28)	(0.28)	(0.18)	(0.03)	(−0.21)
Δ Household debt t-1 * postwar				0.00	0.03
				(0.05)	(0.42)
Δ Household debt t-1 * post-1980				0.02	0.06
				(0.42)	(1.26)
Δ Household debt t-2	0.00	−0.00	0.00	0.00	−0.00
	(0.18)	(−0.04)	(0.01)	(0.04)	(−0.21)
1.postwdummy		−0.00	−0.00	−0.00	−0.01
		(−0.53)	(−0.39)	(−0.17)	(−0.67)
1.neolibdummy		−0.00	−0.00	−0.00	−0.00
		(−0.20)	(−0.68)	(−0.80)	(−0.06)
Constant	−0.01*	−0.00	−0.00	−0.00	−0.00
	(−1.90)	(−0.64)	(−0.79)	(−0.85)	(−0.53)
AIC	−526	−539	−520	−519	−533
Durbin-Watson d	1.95	2.04	1.92	1.95	2.00
Observations	112	112	112	112	112
Adjusted R-squared	0.72	0.76	0.71	0.71	0.75
t-statistics in parentheses

***p < 0.01, **p < 0.05, *p < 0.1

In the previous literature, Chen (2006) for the 1980–2004 period with quarterly data finds significant effects of housing wealth on consumption (0.11) in the long run, but not in the short run. We do not find such effects. Berg, and Bergström, (1995) found for 1970–1992 that financial wealth was ‘crucial’ in explaining consumption and that household debt mattered in the short run. However, they show that the results are driven by the period after the financial deregulation of 1985, after which it became easier to borrow against the house to increase consumption. Their short-run coefficients for financial wealth are 0.04 and 0.05, while that of housing wealth are 0.08 and 0.12, depending on model specification. Our results indicate that over the long run there were no such effects in Sweden.

Table 8 reports results for investment for Sweden. ECM specifications did not pass the critical values for cointegration, so we run FD models. We find strong demand effects on investment. Share price growth increase investment by about 0.05. Business debt growth has a large and positive effect on investment, aboyt 0.40, and this effect seems to be especially driven by post-1980 effects. House price growth only increases investment after 1980.

Table 8. Wealth effects on investment in Sweden.

	(1)	(2)	(3)	(4)	(5)
Δ Y t	1.54***	1.50***	1.52***	1.55***	1.42***
	(6.90)	(5.97)	(6.19)	(6.24)	(5.79)
Δ Y t-1	0.51***	0.40**	0.45**	0.51***	0.35*
	(2.70)	(2.05)	(2.36)	(2.69)	(1.81)
Δ Y t-2	−0.77***	−0.77***	−0.78***	−0.86***	−0.89***
	(−4.08)	(−3.88)	(−3.98)	(−4.12)	(−4.41)
Δ Interest rate t	−0.00	−0.00	−0.00	−0.00*	−0.00*
	(−1.59)	(−1.55)	(−1.52)	(−1.97)	(−1.91)
Δ Interest rate t-1	−0.00**	−0.00**	−0.00**	−0.00**	−0.00**
	(−2.26)	(−2.16)	(−1.99)	(−2.60)	(−2.26)
Δ Stock prices t-1	0.05*	0.17***	0.06**	0.04	0.18***
	(1.71)	(2.76)	(2.09)	(1.38)	(3.03)
Δ Stock prices t-1 * postwar		−0.15			−0.15*
		(−1.64)			(−1.67)
Δ Stock prices t-1 * post-1980		−0.15**			−0.16**
		(−2.23)			(−2.38)
Δ House prices t-1	−0.08	−0.07	−0.20*	−0.12	−0.16
	(−0.96)	(−0.76)	(−1.79)	(−1.23)	(−1.47)
Δ House prices t-1 * postwar			0.05		−0.00
			(0.29)		(−0.01)
Δ House prices t-1 * post-1980			0.46***		0.33*
			(2.64)		(1.79)
Δ House prices t-2	−0.11	−0.12	−0.10	−0.14	−0.14
	(−1.35)	(−1.46)	(−1.14)	(−1.57)	(−1.66)
Δ Business debt t	0.46***	0.43***	0.43***	0.25	0.21
	(3.41)	(3.11)	(3.20)	(1.44)	(1.19)
Δ Business debt t * postwar				0.18	0.31
				(0.51)	(0.71)
Δ Business debt t * post-1980				0.42**	0.37*
				(2.05)	(1.74)
Δ Business debt t-1	−0.38***	−0.30*	−0.41***	−0.31**	−0.30*
	(−2.66)	(−1.98)	(−2.74)	(−2.04)	(−1.93)
Postwar dummy		−0.01	−0.01	−0.02	−0.03
		(−0.71)	(−0.45)	(−0.57)	(−0.66)
Post-1980 dummy		−0.01	−0.03*	−0.03*	−0.05**
		(−0.53)	(−1.85)	(−1.75)	(−2.49)
Constant	0.00	0.01	0.01	0.02	0.03*
	(0.26)	(0.73)	(1.07)	(1.18)	(1.99)
AIC	−334	−333	−335	−331	−336
Durbin-Watson d	2.04	2.16	2.15	2.13	2.29
Observations	114	114	114	114	114
Adjusted R-squared	0.69	0.69	0.70	0.69	0.71

t-statistics in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. Year dummies for 1915, 1916, 1919, 1940, 1945 and 1946 not reported.

5. Conclusions

The paper has investigated the effect of financial variables on consumption and investment for four economies with long historical data. Table 9 gives a summary of the key findings. We find economically signficant effects for all countries, but statistically significance varies and is often sensitive to the specification. Comparison of results is complicated by the fact that we used ECM specifications for France, but had to use difference specifications for UK, Norway and Sweden. We find that stock prices have short run positive effects on consumption in two of our countries: +0.05 in Britain and +0.02 in Norway. In France, we also found an effect, but only after 1945: in the postwar period there we also estimated an effect of +0.05. For these countries, the result is very similar to Modigliani’s old five cents to the dollar rule, as well as recent empirical research (Ludvigson and Steindel 1999; Ludwig and Slok 2004; Fair 2017). As we could have expected, as housing assets are more widespread in the population than stocks (cf. Ludwig and Slok 2004), the effects of house prices on consumption are larger than those of stock prices. In Britain and Norway, the effects are twice as large as the stock price effects, at +0.10, and there is also a positive effect in France, at 0.06. Household credit expansion increases consumption in Britain (+0.05) and France (+0.15). We find little direct support for the suggestions (Berg, Bergström, and Bergstrom 1995; Lindner 2014) that wealth effects on consumption have become bigger since the deregulations of the 1980 s; only that stock prices have stronger effects after 1980 in France, and house prices in Britain. But the rather consistent effects of household credit points to the connection between housing values and the availability of household credit as an important factor (as in Slacalek 2009; Mian and Sufi 2014).

Table 9. Overview of results.

	Preferred Specification: table, column, model type	Main results	Period-specific effects
Consumption
UK	1.5 Diff	Positive effects of PS (~0.05), PH (~0.10) and DHH (~0.05)	PS effects stronger before 1945 PH effects stronger after 1945 and especially after 1980
France	3.5 ECM	Long run positive effect of PH (0.06) Almost significant short run positive effect of DHH (~0.05)	PS has positive effect after 1945 (0.05), strongest after 1980 (0.10)
Norw	5.5 Diff	PS positive (0.02) PH positive (0.10)	DHH negative effect postwar PS slightly negative effect post-1980
Swe	7.5 diff	Short-run negative effects of PS (−0.03)
Investment
UK	2.5 diff	Interest effect negative PS effects positive [~0.10) PH effects negative (~-0.30)	Negative effect of PS after 1980
France	4.5 ECM	Short-run PS positive (~0.15) Short-run D positive (0.28)	No time-varying effects
Norw	6.5 diff	PS positive (~0.25)	Weaker effect of PS after 1980 (~0.05)
Swe	8.5 diff	PS short-run positive (~0.10) DBUS short-run positive (~0–30)	PS no effect after 1980 PH positive short-run effect after 1980 (~0.20) DBUS short-run effect strongest after 1980

As for investment, in all four countries it is boosted by stock prices. The effect varies from +0.10 in Britain, +0.25 in France, +0.15 in Norway and +0.20 in Sweden. House prices have no positive effect on investment, while credit expansion boosts investment in France (0.30) and Sweden (0.30).

What about a changing role of financial variables for investment over time? In Sweden, stock prices seem to have had a positive effect on investment before 1945, but not in the postwar or post-1980 periods. Furthermore, there is a very strong positive effect of housing prices on investment after 1980, but not in the postwar period. This indicates a more housing-driven investment cycle after 1980, which chimes well with the booming house prices in this period (cf. Knoll, Schularick, and Steger 2017,; Gustafsson, Stockhammar, and Österholm 2016 on Sweden). On the other hand, in Britain and Norway, there is less of a significant association between stock prices and investment after 1980. This can possibly be interpreted as an indication of a weaker association between profitability and investment in the period of ‘financialization’.

Wealth effects differ between asset classes, over time and between countries. But how? Firstly, we find that while share prices boost investment in all four countries, we find less consistent positive wealth effects on consumption. For consumption one might say that we find support for the idea that the more market-based economy of Britain has larger wealth effects than the more ‘coordinated’ market economies of France, Norway and Sweden. Britain is the only country in our study where we find significant effects on consumption of all financial variables – stock prices, house prices, and household debt. This is in line with a literature showing larger wealth effects on consumption in Anglo-Saxon countries (e.g. De Bonis and Silvestrini 2012; Ludwig and Slok 2004; Slacalek 2009). In contrast we find relatively similar effects of stock prices on investment (ranging from 0.10 to 0.25) in all four countries.

Our investigation demonstrates a possible route forward for macroeconomic research on wealth effects, in that it highlights the need to go beyond the post-1970 s period researched in a large literature. Interest in wealth effects has often been stimulated by specific episodes such as the stock market losses in 1973–75 (Bosworth, Hymans, and Modigliani 1975) or the US stock market boom of the 1990 s (Poterba and Samwick 1995). And of course, the interplay between booms and busts took place also in earlier periods. In this paper, the emphasis has been on exploring wealth effects over the long run. In further research, one could deepen our understanding of the macroeconomy and the interactions between the financial sector and the real economy, by looking more specifically at periods of turbulence, for example, zooming in on the effects of stock and housing market turbulence (and increased regulation) in the 1920 s and 1930 s. The data compiled in the present paper do show on average more variation in the post-1980 period than in the postwar period, but there are significant events of up- and downturns earlier too. Here, one could also explore further the finding from this paper, of stronger and earlier wealth effects in Britain than in France, Norway and Sweden. Since the British financial markets were earlier well-developed (cf. Acheson, Campbell, and Turner 2017; Rutterford et al. 2011), does this also mean that for instance the recession of the early 1920 s was propagated to a larger extent through wealth effects channels in Britain? Such more specific investigations are outside the scope of the present paper, but we believe that our paper points in this direction for further interesting research on wealth effects.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the Institute for New Economic Thinking [INO1500007].

Notes

1. In the 1990 s, after financial deregulation-induced credit booms, conventional models of consumption broke down (Eitrheim, Jansen, and Nymoen 2002, 40). Conversely, the large drop in stock market values in 1973–75 stimulated discussion of its depressive effects on consumption (Bosworth, Hymans, and Modigliani 1975).

2. Even though there is debate on whether the country in the 1990 s moved to a market-based system (Bertrero 1994).

3. There is a discussion of whether the estimated effects are causal or not. Poterba and Samwick (1995) argue that financial effects (i.e. rising stock prices quarter t correlated with more investment quarter t + 1) are really leading indicators effects. Dynan and Maki (2001) use US consumer surveys to investigate the issue and draw the opposite conclusion, that wealth effects (out of stock prices) are real and not just a spurious correlation.

4. Campbell and Cocco (2007) analyse the UK and find great heterogeneity in the effect of rising house prices, with older house owners consuming more, while young renters are not affected at all. On the aggregate, the question is of course the size and importance of the older house owners’ effect.

5. See also the discussion in Davis (2010). Chirinko (1993) provides a good, if dated, overall discussion of investment and its determinants.

6. The marginal propensity to consume varies by income source, so for example Davis and Palumbo (2001, 33) find that it is lower for capital incomes than for other incomes. For this reason, in robustness checks (not reported) we include the wage share, which measures the share of national income which accrues to wage earners.The results regarding wealth effects do not change.

7. Note that the standard critical values do not apply to error correction coefficient. Banerjee, Dolado, and Mestre (1998, Table 1) report that a with T = 100 and 5 regressors, the critical value is – 3.85 for the 90% level.

8. Goodhart and Hofmann (2008), who use quarterly data, use four lags. Kothari, Lewellen, and Warner (2016, 13) who use quarterly data for the US claim that the effects of profits and stock prices on investment play out in up to a year and a half.

9. Since these specifications only encompass the years 1958–2016, it does not make sense to include both postwar and post-1980 dummies. Thus, we only include the post-1980 one.

10. For a study of wealth effects on consumption in France during the 2008–09 crisis, see Arrondel, Savignac, and Tracol (2014).

11. The SCB’s inflation index for 1831–2018 is available at https://www.scb.se/hitta-statistik/statistik-efter-amne/priser-och-konsumtion/konsumentprisindex/konsumentprisindex-kpi/pong/tabell-och-diagram/konsumentprisindex-kpi/inflation-i-sverige/(accessed 11 November 2019).

12. https://www.scb.se/hitta-statistik/statistik-efter-amne/boende-byggande-och-bebyggelse/fastighetspriser-och-lagfarter/fastighetspriser-och-lagfarter/pong/tabell-och-diagram/fastighetsprisindex-ar-1981100/(accessed 11 November 2019).

13. http://www.nasdaqomxnordic.com/index/historiska_kurser?languageId=3&Instrument=SE0000337842 (accessed 11 November 2019).

14. Available at https://www.scb.se/hitta-statistik/statistik-efter-amne/nationalrakenskaper/nationalrakenskaper/nationalrakenskaper-kvartals-och-arsberakningar/(accessed 11 November 2019).

15. Available at https://www.scb.se/hitta-statistik/statistik-efter-amne/finansmarknad/finansrakenskaper/finansrakenskaper-kvartal-och-ar/(accessed 11 November 2019).

Appendix

For Britain we get GDP data, interest rates, debt data, the wage share, house prices, stock prices and CPI all from the Bank of England dataset ‘Three centuries of British Macroeconomic Data’ (Hills, Thomas, and Dimsdale 2010).

For France we use Piketty and Zucman’s (2014) compilation of historical national accounts, and take financial series from the cross-national studies of Jordà, Schularick, and Taylor (2016b) and Knoll, Schularick, and Steger (2017). We have linearly interpolated the real exchange rate 1915–18 and the interest rate 1915–21. Other than that, the series that we use are quite coherent and complete. The national accounts data from Piketty & Zucman have been completed for the years 2011–2016 with data from the official National Accounts from insee.fr. To facilitate linking, we have simply spliced the series by using the growth rates from the NAs and applying them to the P&Z data from 2010. To extend the stock price index beyond 2013, we have used data from OECD for the years 2014–2016 (https://data.oecd.org/price/share-prices.htm). National accounts data are missing for the years 1940–1948. For France business debt is only available since 1958, and for Norway since 1978. Then we use (a) mortgages and (b) total loans to non-financial sector since 1900, and calculate a proxy for business debt as (b) less (a). For France, the correlation between the thus proxied household and business debt stocks, and the actual data from 1958 on, is 0.997 for both household and business, so we reason that these proxies are reasonable.

For Norway we get data are mostly from two massive books on historical national accounts: GDP and its components from Grytten (2004a), the wage share from Bengtsson and Waldenström (2018), interest rates from Eitrheim and Klovland (2007), debt, house prices and share prices from the Jordà, Schularick, and Taylor (2016b) database, and CPI from Grytten (2004b). For Norway share price data are only available since 1914 (cf. Klovland 2004), which limits our analysis to starting in this year.

For Sweden we take GDP and its components from Krantz and Schön (2015), the wage share from Edvinsson (2005) supplemented by AMECO data after 2000, interest rates from Waldenström (2014), debt from Ahnland (2015), house prices from Söderberg, Blöndal, and Edvinsson (2014), share prices from Waldenström (2014), and CPI from Statistics Sweden’s long-run series.¹¹ Ahnland’s debt data separates between mortgages and other types of credit. The house price series has been updated to 2016 using data from Statistics Sweden.¹² The stock price series has been updated using data from the bourse itself, Nasdaq OMX.¹³ GDP data has been updated using data from Statistics Sweden, the data release ‘GDP: expenditure approach 1950–2017, aggregated, volume index year 2015 = 100’.¹⁴ The debt series are updated for 2014–16 using data on household borrowing and total borrowing from Statistics Sweden, category Finansräkenskaper (Financial accounts).¹⁵

The datasets are available from the authors in Excel or Stata format.