The Puzzle with Increasing Money Demand – Evidence from a Cross-Section of Countries

Graf Lambsdorff, Johann Article The Puzzle with Increasing Money Demand – Evidence from a Cross-Section of Countries Kredit und Kapital Provided in Cooperation with: Duncker & Humblot, Berlin Suggested Citation: Graf Lambsdorff, Johann (2005) : The Puzzle with Increasing Money Demand – Evidence from a Cross-Section of Countries, Kredit und Kapital, ISSN 1865-5734, Duncker & Humblot, Berlin, Vol. 38, Iss. 2, pp. 155-176, https://doi.org/10.3790/ccm.38.2.155 This Version is available at: https://hdl.handle.net/10419/293522 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ Kredit und Kapital, 38. Jahrgang, Heft 2 Seiten 155-176 The Puzzle with Increasing Money Demand — Evidence from a Cross-Section of Countries By Johann Graf Lambsdorff1, Passau I. Introduction The positive impact of income on the ratio of money demand to GDP is standard to empirical investigations - and so is its contradiction with theory. Since Baumol, Tobin and Allais one expects the ratio of money demand to GDP to decrease with income, Baumol and Tobin (1989). The higher the transaction volume (proxied by income) the more one will endeavor to economize on monetary holdings. That this relationship is not given empirical recognition is precarious. This inconsistency has been commonly related to structural shifts arising in the long run. Our theoretical predictions would be valid only for short-term cyclical fluctuations. Friedman (1959) pioneered the assumption of a long-term upward trend of money demand (that is, a downward secular trend of income velocity). If this assumption holds, one should be able to identify the causes of structural changes and attempt to provide quantitative estimates. This study shows reproduces the standard empirical finding of a positive association between income and the ratio of money demand to GDP. But it shows that this association can be traced to the correlation of income with other variables: The share of agriculture, life expectancy at birth, openness, and trust in the banking system. Including these variables reconciles theory with evidence: A negative impact of income on money demand (relative to GDP) is obtained. The study is organized as follows. Section 2 provides theoretical arguments for a long-term trend to increasing money demand (relative to GDP). Some basic regressions are carried out in Annex 2. These are in1 The author is grateful to J. Breitung, H.-J. Jarchow, H. Môller, M. Schinke and C. Schinke for providing helpful comments and to the participants of a workshop at the Institute for Statistics and Econometrics, Gôttingen, Germany, January 2003, and a workshop at the economics department, University of Regensburg, June 2004. Kredit und Kapital 2/2005 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.38.2.155 | Generated on 2023-01-16 13:22:12 156 Johann Graf Lambsdorff tended to determine the appropriate functional form of GNP per head and the interest rate. These regressions are carried out for a cross section of 126 countries and bring about the standard results of a negative impact of the interest rate (or a positive impact of the reciprocal term) and a positive impact of income on the ratio of money demand to GDP. These results are not crucial to this study and are rendered to the appendix accordingly. Section 3 introduces some of the crucial variables mentioned. These are added to the regressions and explanations are provided how they contribute to explaining money demand. Section 4 extends the investigation by assessing also portfolio demand for money and including inequality in the analysis. Section 5 concludes. Annex 1 explains how the data was retrieved and compiled. Annex 3 reports the list of countries contributing to the regressions and annex 4 reports the crucial correlations. II. Long-Term Determinants of Money Demand One reason for a secular upward trend of money demand may relate to monetization. Monetization is defined as the fraction of total income received in the form of money (Melitz and Correa (1970, 13); Jones (1976); Niehans (1971)). If, on the other hand, barter trade characterizes societies, exchange can be carried out without making use of money. The higher is the share of transactions based on money, the higher will be the ratio of money demand to the total transaction volume. As societies become richer, there is likely to be a tendency to depart from barter. This could explain a secular trend where growing economies experience an increasing money demand. Another impact on monetization may result if societies are characterized by a widespread use of reciprocal exchange. Economic actors then start to accumulate rights for reciprocal favors instead of piling up financial assets.2 Reciprocity thus becomes an alternative medium of exchange, lowering the necessity to hold money. A variety of variables can be tested that are likely to relate to monetization. Driscoll and Lahiri (1973) and Melitz and Correa (1970, 15), for example, suggest the share of the agricultural sector in national income. In a subsistence economy production and consumption go hand in hand. The basic exchange that takes place might be more of a barter type. The agri2 Kranton (1996) shows how the importance of reciprocity can become self-perpetuating when goods are only available to members of a network. Kredit und Kapital 2/2005 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.38.2.155 | Generated on 2023-01-16 13:22:12 Evidence from a Cross-Section of Countries 157 cultural sector is therefore likely to go along with little monetization and a low money demand.3 Another reason for the secular upward trend emerges when incomes - even those received in the form of money - do not perfectly correlate with the transaction volume. An increasing division of labor characterizes modern societies. This characteristic of technological progress has been dealt with more explicitly in recent models of economic growth. These models emphasize the role of R&D in creating process innovation, allowing a deeper and thus more productive division of labor. I follow Grossman and Helpman (1991, ch. 3) and assume a constant returns to scale CES production function where A intermediate products x(j) are combined to produce the output c: The share Lx of the total labor force L is used for production. The rest is employed for R&D. The intermediate products are produced only with labor input. Assuming input for intermediate products to be equal to output (x(j)) and the labor force to be divided equally for the production of the A intermediate goods, we obtain: x(j) = Lx/A. Inserting this brings about output, c, in relation to the amount of intermediate products that are employed in the production and the productive labor force: 3 Melitz and Correa (1970, 15) also test the urbanization ratio as a proxy for monetization. In accordance with their results I did not find a significant impact of this variable and refrain from reporting the details. The level of corruption in a country may also serve as a good proxy for reciprocity. Reciprocity can induce bureaucrats to serve their kin or members of a network instead of being devoted to the public. This can be the source of corruption. Likewise, for corruption to flourish trusted relationships must be established, which keep partners in the corrupt relationship from acting opportunistically Countries with an established reciprocal exchange therefore provide ground for corruption, whereas a high level of monetization makes it difficult to strike a corrupt deal (Lambsdorff (2002)). Yet, corruption did not exert a significant influence on money demand. Exchanging goods with the help of money may require mathematical skills, suggesting that monetization increases with literacy. However, also this variable was insignificant. Modern capitalism has been traced to Protestant ethics. One may assume that optimizing money demand is particularly an issue in countries with a large share of Protestants. Again, the share of Protestants was tested and found to be insignificant. c = A{1~a)/aLXKredit und Kapital 2/2005 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.38.2.155 | Generated on 2023-01-16 13:22:12 158 Johann Graf Lambsdorff Thus, production increases with the labor force and the amount of intermediate products A. This latter term depicts the positive (but less than proportional) impact of the division of labor on production. Endogenous growth can now result when the rest of the labor force, LA = L - Lx, is employed in R&D, and when there are positive spillovereffects such that knowledge increases proportional to A. This suggests that A increases constantly over time. Employing this model for our purposes, it is straightforward to observe that the transaction volume in the factor market is depicted by ALX. As a result, increases in A proportionately increase the transaction volume but bring about a less than proportional increase of output. Technological progress, as depicted by this model, will therefore increase the ratio of money demand to GDP, forcing the aforementioned secular trend.4 The extent of technical know-how with regard to process innovation is difficult to quantify, particularly for a cross-section of countries. I suggest that life expectancy at birth captures a good deal of this variable.5 The best form of storing knowledge might be human brain. In the case of learning-by-doing one cannot store knowledge elsewhere. Societies with a higher life expectancy would then provide a more economic usage (that is, a lower depreciation rate) of this knowledge. I will therefore include life expectancy at birth as an explanatory variable. Specialized knowledge that is necessary for process innovation may also be vulnerable to a holdup - an argument that is standard to the New Institutional Economics. Long and repeated exchange between firms and the producers of intermediate inputs can help to establish reputation effects or contain opportunism with the help of repetition. Higher life expectancy allows the exchange to become more durable, letting repetition become more likely and investments into specialized knowledge less likely to be exploited by opportunism.6 4 In line with this argument, for the last decades world trade grew substantially, indicating a worldwide trend to an increasing division of labor. At the same time world GDP grew too, yet at a slower pace. Thus, the ratio of (cross-border) exchange to output increased. 5 That the aging process can have an impact on money demand has been recently given recognition {Remsperger (2001)). 6 Other indicators of human capital instead of life expectancy at birth at may be considered for inclusion, for example school enrollment rates. These depict investments into human capital, but they refer less to a type of knowledge that is obtained by learning-by-doing. To the extent that process innovation is acquired on the job, they do not well depict this type of knowledge. Assessments of "years in school" were insignificant in the regressions. Kredit und Kapital 2/2005 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.38.2.155 | Generated on 2023-01-16 13:22:12 Evidence from a Cross-Section of Countries 159 Countries that are well integrated into the world economy are also likely to be characterized by a deeper division of labor. If countries are strong in exchanging intermediary goods with other countries, they are embedded in a global division of labor. Such countries will have a high degree of openness (the ratio of export plus import to GDP), a variable that will be included in the regressions.7 Portfolio demand for money, finally, may overshadow transaction demand. Moller and Jarchow (1996) argue that the existence of portfolio demand is the reason for a decreasing secular trend and that its incorporation reconciles theory with evidence. They argue for the case of Germany that transaction demand for money increases at a lower pace than GDP, in line with theory. Money is also held for portfolio considerations and the increasing wealth accounts for an additional increase in money demand. It is this effect that they hold responsible for the secular trend. In regressions that do not control for wealth, the impact of income (which correlates with wealth) will therefore be biased, misrepresenting our theory. Whether the coefficient is biased upward or downward, however, depends on whether the ratio of wealth to GDP increases or decreases over time. A neoclassical growth model commonly suggests that the ratio of capital to GDP remains constant because in a steady-state both variables equally grow with population and with technological progress. Yet, one reason for wealth increasing relative to GDP arises with the aging of society. Following a life-cycle hypothesis of consumption, the savings needed after retirement increase with life expectancy. This changes the ratio between wealth and income and increases money demand relative to GDP due to increased portfolio demand. Holding money may also be preferred to other forms of financial capital. A crucial impact can result from an efficient and trusted banking system. Demand, time, and savings deposits, for example, are likely to 7 This measure of openness is somewhat distorted by a country's population, with larger countries being characterized by a lower openness. The larger a country, the more of its trade is domestic and not with foreign countries. For example, the total EU's openness may not surpass that of the USA, but the values for each of its member countries surely do. This means that trade within the EU is included in the statistics while trade between Texas and California is not. This distortion can be corrected by including also population into the regression. Population may also impact on currency substitution. The more likely it is that a randomly selected trading partner uses the same currency, the less interest people have in holding foreign currencies. Incorporating population into the regressions brought about insignificant results and did not contribute to our analysis. Kredit und Kapital 2/2005 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.38.2.155 | Generated on 2023-01-16 13:22:12 160 Johann Graf Lambsdorff increase with the development of honest banks and will remain poor if no trusted banking system is in place. Once these deposits are secure, money becomes attractive relative to holding other assets. Yet, there is hardly convincing and sufficient data on trust in banks available for a cross-section of countries. What might serve as a proxy for our purpose is the ratio of currency to the total money demand (Hanson and Vogel (1973, 366)). An underdeveloped banking system suggests that money must be held in the form of currency. To the contrary, less currency is needed if money can be held in banking accounts. To the extent that people trust banks they will substitute currency for deposits (Claque et al. (1999, 188)). At the same time they are more likely to hold deposits instead of other non-monetary financial assets. For example, savings deposits become the more attractive as compared to stocks the more people trust the banking system. This suggests that a low ratio of currency to money is likely to go along with an overall higher money demand.8 The ratio of currency to money will also have an impact on money supply. Private banks can increase lending and deposits when little currency is demanded from their clients. Increased money supply lowers interest rates and thus also increases money demand. However, this argument alongside with the recognition that money supply might be positively affected by the interest rate, suggests than the interest rate might be endogenous to the model - requiring adequate subsequent statistical treatment of this variable. III. Empirical Evidence Time series analysis is the standard approach to investigating money demand. There only exists scattered evidence on money demand in a cross-section analysis. Noteworthy but dated exceptions are Doblin (1951), Ezekiel and Adekunle (1969), Melitz and Correa (1970) with a reply by Wallich (1971) and Lothian (1976). A more recent one, which, unfortunately, disregards important explanatory variables, is provided by Sell (1997). In a related investigation, Lucas (1996, 665) correlates inflation and monetary growth for a cross-section of countries.9 8 Melitz and Correa (1970, 13-4) also include the ratio of currency to money as an explanatory variable. They provide an additional argument for its likely impact. Because currency holdings entail substantially higher risks than demand deposits people will more strenuously try to economize on money. Kredit und Kapital 2/2005 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.38.2.155 | Generated on 2023-01-16 13:22:12 Evidence from a Cross-Section of Countries 161 A cross-section approach can enrich our understanding of the explanatory variables because it allows for the inclusion of variables that show little variation over time or that are even unavailable as time series. It therefore makes sense to validate results from time series analysis with those from cross-sections, particularly with regard to the long run forces governing the demand for money (Wallich (1971), Lucas (1996, 665)). I determine money demand for a representative individual in each country. The ratio of money to income, our dependent variable, is independent of whether we look at individual demand or that of the country at large. For our dependent variables, however, it is important to keep in mind that an individual's money demand is modeled. This approach was superior in avoiding a country's size to exert an impact on the regressions. The ratio of money to income is both a real and a nominal ratio, because the price level for both variables would be the same. I determine an individual's real money demand as a function of real income per head and a variety of further explanatory variables. Our dependent variable is the ratio of money demand to (nominal) GDP, while our explanatory variables are (ppp adjusted) GNP10 per head and the interest rate. This brings about the standard results, as shown in table 1, regression 1. This regression repeats the results from regression 5, table A.2, annex 2, where the choice of the functional form has been derived. While GNP per head increases money demand relative to GDP (in contrast to the standard model by Baumol and Tobin), the interest rate lowers it. Instead of the simple (logarithm) of the interest rate, its reciprocal term has been used, which subsequently exhibits a positive impact. Details on the choice of the functional form are provided in annex 2. Introducing the four aforementioned variables brings about significant results. For all four variables I obtain the expected impact at a high significance level. An increase in agriculture from 10 to 20 percent reduces money demand by 18 percent. An increase of life expectancy by one year increases money demand by 1.9 percent. A doubling of the ratio of currency to money reduces money demand by 26 percent.11 Increasing open9 Panel-data, certainly, combine time-series and cross-section data to a richer investigation of money demand. This approach is provided in Herwartz and Reimers (2001). 10 I chose to use GNP rather than GDP because income is likely to better explain money demand than production. 11 One might consider the existence of a direct impact of the ratio of currency to money on the ratio of money to GDP: Increasing money while keeping all other variables constant suggests a direct elasticity of -1. But this argument deserves credibility only if we can find a variable that changes money demand but leaves Kredit und Kapital 2/2005 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.38.2.155 | Generated on 2023-01-16 13:22:12 162 Johann Graf Lambsdorff ness by 10 percentage points increases money demand by 2 percent. Once including the share of agriculture or life expectancy the impact of GNP per head falls to insignificance. The share of agriculture and life expectancy thus outperform income as a cause of the secular trend. If all variables are included simultaneously, as in regression 6, table 3, GNP per head enters even with a significant negative coefficient. The results are checked in regression 6 by including regional dummies. This is to make sure that the results are of global validity and not driven by regional peculiarities. Yet, these variables are largely insignificant. The various regressions have also been tested for linearity. Running a White-heteroskedasticity test provides no evidence for rejecting the simple linear form employed. As shown in regression 7, including a squared term for GNP per head did not affect the significance of the four variables. The impact of the four variables is therefore not related to the idea that an inappropriate functional form of GNP per head was employed. Altogether, the share of agriculture, life expectancy, the ratio of currency to money, and openness are important variables, which deserve recognition for the explanation of money demand. The interest rate and money demand interact in a more complex way when also considering money supply effects. The interest rate may not be exogenous to money demand but balance money demand and supply, where banks increase their lending and the money supply multiplier in response to higher interest rates. Testing the relevance of this effect for our regressions requires an instrument that well correlates with the interest rate but not with the error term of the model. The (logarithm of the) inflation rate well serves this purpose. It affects money demand the other variables unaffected. There is no apparent candidate for this effect. Thus, the idea of keeping the other variables constant must be considered implausible. In a similar spirit we would not reject the findings by Feldstein and Horioka on the mere grounds that the ratio of investment to GDP and savings to GDP would be driven by variations of GDP - in this case the idea of holding investment and savings constant would be equally implausible. A plausible impact of the ratio of currency to money on the ratio of money to GDP is the one related to trust in the banking system. A further impact may run via money supply, because a low ratio of currency to money allows banks to increase the money supply. But given equilibrium in the money market, this increases money demand only if the interest rate decreases. Given that we control for the interest rate, this impact is already captured in the model. Also changes in reserve money would have an impact on the interest rate that affects money demand only via changes in the interest rate. I checked for the significance of reserve money and found that it becomes insignificant once the interest rate is included in the regression - in line with our expectations. Kredit und Kapital 2/2005 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.38.2.155 | Generated on 2023-01-16 13:22:12 Evidence from a Cross-Section of Countries 169 Data on openness are from the World Penn Tables and reflect average values between 1974 and 1992. Data on the share of Protestants and (ppp adjusted) GNP per head are from LaPorta et al. (1999). The share of agriculture as percent of GDP, the share of literates, and life expectancy at birth are taken from the CIA Factbook 2000. The share of urban population in 2000 is determined by the United Nation's Statistics Division. The 2001 Corruption Perceptions Index is from Transparency International. Inequality in income is measured by the Gini-coefficient. The data are from the World Development Indicators 2002 and were compiled largely between 1991 and 1998. Some missing countries have been substituted by assessments from Deininger and Squire (1996).14 A country's savings stock was determined first by using gross national savings data from the World Savings Data Base, a project by the World Bank.15 I determine the average over the years 1974-1995. More recent data was not available. From this I determine the ratio of the savings stock to GNP by a perpetual inventory method, assuming an annual depreciation rate of 7 percent.16 This is done here by computing a truncated savings stock (S[995), resulting from the gross savings between 1974 and 1995. To this truncated stock I add the savings stock from 1973: S1995 = S'1995 + (1.07)"22Si973This old stock (S1973) is determined by introducing the assumption that the ratio of accumulated savings to GDP remains constant over time.17 The old savings stock is then given by S1973 = GNP1913 • S1995/GiVPi995. Inserting this and solving for S1995 produces: g ^ 1995 1995 ~ 1 - (1.07)~22GiVP1973/GiVP1995 ' This procedure was particularly helpful in adequately dealing with countries whose savings data was incomplete. For example, if savings data prior to 1985 was not available, I would add (1.07)-11Si984 to S'1995. 14 The authors provide updated data at the following address: http://www. worldbank.org/research/growth/dddeisqu.htm is The data can be retrieved from http://www.worldbank.org/research/projects/ savings/data.htm 16 The value of 7 percent is considered to be a realistic approximation of the real world depreciation rate, see for example Benhabib and Spiegel (1994). I? This assumption is in line with the steady-state in neoclassical growth models but in contrast to ideas that wealth might increase more than GDP. But even if it does, this impact is likely to be small. Using the procedure presented here, the ratio of GDP to capital has also been determined for 1990. It was observed that the value on average for all countries hardly differed from the one for 1995. Kredit und Kapital 2/2005 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.38.2.155 | Generated on 2023-01-16 13:22:12 170 Johann Graf Lambsdorff Since this term is larger than (1.07)-22 Si973 lack of savings data does not result in a smaller assessment of the savings stock. From the resulting savings stock I determine a ratio of the savings stock to GDP (Sl995/GDP1995). Annex 2: Preparatory Regressions There are 126 countries with data on inflation, GNP per head, and monetary aggregates. As a dependent variable I employ the ratio of a broad monetary aggregate (money + quasi money) to GDP, that is, the inverse of income velocity. Regression 1, table A.l, reports the results. Inflation diminishes money demand while GNP per head exerts a positive impact on money demand. A doubling of income increases the ratio of money demand to GDP by more than 22 percent. This value seems to be largely in line with evidence from time series (Fair (1987)). A standard impact on money demand can be expected from interest rates because these indicate the foregone profit from holding money visà-vis investing into other interest bearing assets. I therefore include the deposit rate and lending rate in regressions 2 and 3. As shown in regression 4, the lending rate exerts the strongest influence. It will be used henceforth. In order to carry out subsequent regressions for as large a coverage as possible, I determine missing values of the lending rate by their forecasted value. I carry out this forecast by regressing the lending rate on the deposit rate (if not available on the inflation rate) and determine forecasted values from this regression. The correlation coefficient between inflation and the interest rates is above 0.8; that between the two interest rates is close to 0.9, suggesting that the forecast should be good. This extended index for the lending rate is used in regression 5. It obtains a negative impact at a high significance level. The 126 countries included in regressions 5 are reported in Annex 1. Various tests have been carried out to find out whether the impact of the (logarithm of the) interest rate is linear, indicating a constant elasticity. A White heteroskedasticity test indicates problems with regard to linearity. In table A.2, regressions 1 and 2, the sample of countries is divided into those with low and those with high interest rates. In line with the idea of a liquidity trap, the coefficient for the interest rate is much larger for countries with low interest rates.18 This suggests that alternative functional forms should be used for regressions with the complete Kredit und Kapital 2/2005 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.38.2.155 | Generated on 2023-01-16 13:22:12 Evidence from a Cross-Section of Countries 171 Table A.l OLS Dependent Variable: (Logarithm of) the Ratio of M2 (Money + Quasi Money) to GDP Independent variables 1 2 3 4 5 Constant 2.30 (9.3)#) 3.03 (7.5) 4.09 (5.3) 3.89 (4.4) 3.99 (7.5) GNP per head (log.) 0.222 (8.3) 0.241 (9.1) 0.212 (7.1) 0.228 (7.9) 0.210 (8.2) Inflation (log.) -0.122 (-3.2) -0.000 (-0.0) Deposit Rate (log.) -0.173 (-4.0) 0.123 (0.9) Lending Rate (log.) -0.272 (-3.1) -0.375 (-2.0) Lending Rate (log.), extended** -0.256 (-4.3) Obs. 119 115 104 96 126 Adj. R2 0.46 0.47 0.47 0.51 0.45 JB+) 2.1 3.3 1.6 4.8 1.7 #) All t-statistics (given in parenthesis) are White-corrected to adjust for heteroskedasticity. *) For missing values a forecast has been determined by the deposit rate and the inflation rate. +) The Jarque-Bera measures whether a series is normally distributed by considering its skewness and kurtosis. The assumption of a normal distribution can be clearly rejected for levels above 6. is This result contradicts the idea that the interest rates impacts on money demand in the form of a semi-elasticity. Fair (1987, 473), for example, rejects the idea of a constant elasticity and argues that an increase of the interest rate from 5 to 6 percent should have a smaller impact (on money demand per head) than an increase from 50 to 60 percent. This leads him, as many others, to include the interest rate in a non-logarithmic form, where the coefficient depicts a semi-elasticity. But quite to the contrary, our results suggest that an increase of the interest rate from 5 to 6 percent has an even greater impact than one from 50 to 60 percent. I also ran regressions with the non-logarithmic form, but the non-logarithmic interest rates show a highly uneven distribution with some countries being outliers. The assumption of linearity could be rejected for the regressions. Kredit und Kapital 2/2005 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.38.2.155 | Generated on 2023-01-16 13:22:12 172 Johann Graf Lambsdorff Table A.2 OLS Dependent Variable: (Logarithm of) the Ratio of M2 (Money + Quasi Money) to GDP Independent variables 1. Low Interest 2. High Interest 3. Low Interest 4. High Interest 5 6 Coun-Coun-Coun-Countries tries tries tries Constant 7.66 3.44 1.12 0.784 1.33 -1.63 (5.1) (4.9) (2.9) (1.0) (6.3) (-1.2) GNP per head (log.) 0.158 0.256 0.159 0.256 0.194 0.982 (5.7) (3.6) (5.7) (3.6) (7.3) (2.8) GNP per head (log.), squared -0.054 (-2.3) Lending Rate (log.), extended -0.725 (-3.6) -0.224 (-2.1) 1/log (Lending Rate), extended 3.48 (4.0) 2.81 (2.0) 2.34 (4.8) 2.86 (5.0) Obs. 62 64 62 64 126 126 Adj. R2 0.47 0.24 0.48 0.25 0.47 0.49 JB 2.3 0.3 2.4 0.2 2.5 3.0 sample of countries. A variety of functional forms have been tested, the aim being that the assumption of linearity should subsequently hold. It turned out that the inverse of the (logarithm of the) interest rate well serves this purpose. As shown in table A.2, its impact is largely equal for high interest and for low interest countries. This variable will be used from now on. I also tested the impact of an interaction term: GNP per head times the interest rate. Irrespective of the functional form chosen for the interest rate this turned out insignificant; the results are not reported here. An important question arising is whether also the relationship for income is linear. Based on data for five industrial countries for the last 100 years, Bordo and Jonung (1990) argue that initially income velocity drops with increasing income (the ratio of money to GDP increases), but that this impact is reversed as countries become richer and financial inKredit und Kapital 2/2005 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.38.2.155 | Generated on 2023-01-16 13:22:12 Evidence from a Cross-Section of Countries 173 novations provide sufficient substitutes to holding money. This results in a concave curve for the ratio of money demand to GDP, where money demand is negatively affected by increasing income beyond a certain threshold. I run a rather simple test here by including also the quadratic term of GNP per head, thus testing whether a second order polynomial fits the data. As can be seen from regression 6, table A.2, the quadratic term enters significantly into the regression. The coefficients imply that a changing income has no impact on income velocity if countries have a (logarithm of the) GNP per head of 9.1.19 The results by Bordo and Jonung (1990) are thus supported here, although I am applying a totally different set of data. A country such as the United States with a (logarithm of the) GNP per head of 9.6 is located on the decreasing side of the curve. A further increase in GNP is likely to decrease money demand (relative to income) there. For the Euro area, whose member states (which entered separately into the regression) have an average (logarithm of the) GNP per head of 9.2, an increase in income should leave the income velocity largely unaffected. This suggests that the secular trend, at least as induced by a growing income, has come to an end in the Euro area. A regression in section 3 will test whether the quadratic term remains significant once introducing further explanatory variables. Annex 3: List of Countries The 126 countries included are: Antigua, Argentina, Australia, Austria, Bahamas, Bahrain, Barbados, Belgium, Belize, Benin, Bhutan, Bolivia, Botswana, Brazil, Burkina Faso, Burundi, Cameroon, Canada, Central African Rep., Chad, Chile, China, Colombia, Congo, Costa Rica, Cote d'lvoire, Cyprus, Denmark, Dominica, Dominican Rep., Ecuador, Egypt, El Salvador, Equatorial Guinea, Ethiopia, Fiji, Finland, France, Gabon, Gambia, Germany, Ghana, Greece, Grenada, Guatemala, Guinea-Bissau, Guyana, Haiti, Honduras, Hungary, Iceland, India, Indonesia, Iran, Ireland, Israel, Italy, Jamaica, Japan, Jordan, Kenya, Korea, Kuwait, LAO People's Dem. Rep., Lesotho, Libya, Madagascar, Malawi, Malaysia, Mali, Malta, Mauritania, Mauritius, Mexico, Morocco, Mozambique, Nepal, Netherlands, New Zealand, Nicaragua, Niger, Nigeria, Norway, 19 This results from introducing the coefficients for GNP per head (Y) into the first derivative of the dependent variable (M/Y) and setting this equal to zero: d(M/Y)/dY = 0.982 - 2 * 0.054 * Y = 0 Y - 9.1. Kredit und Kapital 2/2005 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.38.2.155 | Generated on 2023-01-16 13:22:12 174 Johann Graf Lambsdorff Oman, Pakistan, Panama, Papua New Guinea, Paraguay, Peru, Philippines, Poland, Portugal, Qatar, Rwanda, Saudi Arabia, Senegal, Seychelles, Sierra Leone, Singapore, South Africa, Spain, Sri Lanka, St. Kitts, St. Lucia, St. Vincent & Grenadines, Sudan, Suriname, Swaziland, Sweden, Switzerland, Syrian Arab Rep., Tanzania, Thailand, Togo, Tonga, Trinidad and Tobago, Tunisia, Turkey, Uganda, United Kingdom, United States, Uruguay, Vanuatu, Venezuela, Zambia, Zimbabwe. Annex 4: Correlations Correlation M2 to GNP Share of Life Ratio of Openness GiniMatrix GDP per head Agri-Expect-Currency (Imp. + Inequaldog.) dog.) culture ancy to Money Exp.)/ ity (Log. ) GDP M2 to GDP (log.) 1.00 0.66 -0.71 0.65 -0.61 0.28 -0.51 GNP per head (log.) 0.66 1.00 -0.84 0.76 -0.76 0.16 -0.44 Share of -0.71 -0.84 1.00 -0.73 0.77 -0.29 0.34 Agriculture Life Expectancy 0.65 0.76 -0.73 1.00 -0.51 0.08 -0.48 Ratio of Currency to Money (Log. ) -0.61 -0.76 0.77 -0.51 1.00 -0.17 0.21 Openness (Imp.+Exp)/GDP 0.28 0.16 -0.29 0.08 -0.17 1.00 0.04 Gini-Inequality -0.51 -0.44 0.34 -0.48 0.21 0.04 1.00 References Baumol, W. J./Tobin, J. (1989): "The Optimal Cash Balance Proposition: Maurice Allais' Priority", Journal of Economic Literature, Vol. 27 (3), 1160-2. - Benhabib, J./Spiegel, M. (1994): "The Role of Human Capital in Economic Development. Evidence from Aggregate Cross-Country Data". Journal of Monetary Economics, 34, 143-73. - Bordo, M. D./Jonung, L. (1990): "The Long-Ron Behavior of Velocity: The Institutional Approach Revisited", Journal of Policy Modeling, Vol. 12 (2), 165-97. - Chandavarkar, A. (1977): "Monetization of Developing Economies", IMF Staff Papers, Vol. 24, 665-721. - Clague, C./Keefer, P[/Knack, S./Olson, M. (1999): "Contract-Intensive Money: Contract Enforcement, Property Rights, and Economic Kredit und Kapital 2/2005 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.38.2.155 | Generated on 2023-01-16 13:22:12 Evidence from a Cross-Section of Countries 175 Performance", Journal of Economic Growth, IV, 185-211. - Deininger, K./Squire, L. (1996): "A New Data Set Measuring Income Inequality", World Bank Economic Review Vol. 10 (3), pp. 565-591. - Doblin, E. M. (1951): "The Ratio of Income to Money Supply: An International Survey", The Review of Economics and Statistics, Vol. 33 (3), 201-13. - Driscoll, M. J./Lahiri, A. K. (1973): "Income-Velocity of Money in Agricultural Developing Economies", Review of Economics and Statistics, Vol. 65 (3), 393-401. - Ezekiel, H./Adekunle, J. (1969): "The Secular and Behavior of Income Velocity: And International Cross Section Study", IMF Staff Papers, Vol. 16 (2), 224-39. - Fair, R. C. (1987): "International Evidence on Demand for Money", Review of Economics and Statistics, Vol. 69 (3), 473-80. - Friedman, M. (1959): "The Demand for Money: Some Theoretical and Empirical Results", The Journal of Political Economy, Vol. 67 (4), 327-51. - Grossman, G. M./Helpman, E. (1991): Innovation and Growth in the Global Economy, Cambdrige: MIT Press. - Hanson, J. S./Vogel, R. C. (1973): "Inflation and Monetary Velocity in Latin America", The Review of Economics and Statistics, Vol. 55 (3), 36570. - Hausman, J. (1978): "Specification Tests in Econometrics", Econometrica, Vol. 46 (6): 1251-71. - Herwartz, H./Reimers, H.-E. (2001): "Long-Run Links Among Money, Prices, and Output: World-Wide Evidence", Discussion paper, Economic Research Centre, Deutsche Bundesbank. 14/01. - Jones, R. A. (1976): "The Origin and Development of Media of Exchange". The Journal of Political Economy, Vol. 84 (4), 757-76. - Kranton, R. E. (1996): "Reciprocal Exchange: A SelfSustaining System", The American Economic Review, Vol. 86 (4), 830-51. - La Porta, R./Lopez-De-Silanes, F./Shleifer, A./Vishny, R. W. (1999): "The Quality of Government", The Journal of Law, Economics and Organization, 40 (1), 222-79. - Lambsdorff, J. Graf (2002): "How Confidence Facilitates Illegal Transactions - An Empirical Approach", American Journal of Economics and Sociology, Vol. 61 (4), 829-53. - Lothian, J. R. (1976): "The Demand for High stuffed his stuffed its fifth of the-Powered Money", The American Economic Review, Vol. 66 (1), 56-68. - Lucas, R. E. (1996): "Nobel Lecture: Monetary Neutrality", The Journal of Political Economy, Vol. 104 (4), 661-82. - Melitz, J./Correa, H. (1970): "International Differences in Income Velocity", The Review of Economics and Statistics, Vol. 52 (1), 12-7. - Möller, H./Jarchow, H.-J. (1996): "Zur Umlaufgeschwindigkeit von M3". Jahrbücher für Nationalökonomie und Statistik, Vol. 215 (3), 308-27. - Niehans, J. (1971): "Money and Barter in General Equilibrium with Transactions Costs", The American Economic Review, Vol. 61 (5), 773-83. - Pindyck, R./Rubinfeld, D. (1991): Econometric Models and Economic Forecasts, New York: Mc Graw-Hill. - Remsperger, H. (2001): "Aging: Is this really a relevant Issue for a Central Bank", Deutsche Bundesbank Spring Conference, Eltville 4 May 2001 and Bank for International Settlement Review, Vol. 36. - Sell, F. L. (1997): "Zu den langfristigen Bestimmungsgründen der Umlaufgeschwindigkeit des Geldes: Ein sektoraler Ansatz", Kredit und Kapital, Vol. 30 (3), 325-47. - Wallich, H. C. (1971): "Income Velocity", The Review of Economics and Statistics, Vol. 53 (2), 200-01. Kredit und Kapital 2/2005 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.38.2.155 | Generated on 2023-01-16 13:22:12 176 Johann Graf Lambsdorff Summary The Puzzle with Increasing Money Demand - Evidence from a Cross-Section of Countries The ratio of money demand to GDP may increase with portfolio demand, monetization, and a deeper division of labor. Using a cross-section approach to money demand for 126 countries this study shows that the share of agriculture, life expectancy at birth, openness, and trust in the banking system capture a good deal of these influences. Once these variables are included, GNP per head negatively impacts on the ratio of money demand to GDP, which is in line with the standard result by Tobin and Baumol. (JEL E41, C21) Zusammenfassung Das Rätsel der steigenden Geldnachfrage - Evidenz aus einer Querschnittsanalyse von Ländern Die Geldnachfrage (relativ zum Inlandsprodukt) erhöht sich bei einem Anstieg des Vermögens, bei einer erhöhten "monetization" und einer vertieften Arbeitsteilung. In einer Querschnittsanalyse von 126 Ländern zeigt diese Studie, dass der Anteil der Landwirtschaft, die Lebenserwartung, der Offenheitsgrad und das Vertrauen in das Bankensystem einen großen Teil dieses Einflusses abbilden. Sobald diese Variablen berücksichtigt werden, ergibt sich ein negativer Einfluss des ProKopf-Einkommens auf die Geldnachfrage (relativ zum Inlandsprodukt). Dies bestätigt die Resultate von Tobin und Baumol. Résumé Le puzzle de la demande croissante de monnaie - Evidence d'une approche transversale entre pays Le ratio demande monétaire/PIB peut augmenter avec la demande de portefeuille, la monétisation et une plus grande division de travail. En utilisant une approche transversale de la demande monétaire entre 126 pays, cette étude montre que l'agriculture, l'espérance de vie à la naissance, l'ouverture et la confiance dans le système bancaire ont une grande influence. Une fois inclues ces variables, le PIB par habitant a un impact négatif sur le ratio demande monétaire/PIB. Ceci s'aligne avec les résultats de Tobin et Baumol. Kredit und Kapital 2/2005 OPEN ACCESS | Licensed under CC BY 4.0 | https://creativecommons.org/about/cclicenses/ DOI https://doi.org/10.3790/ccm.38.2.155 | Generated on 2023-01-16 13:22:12