Barebones Outline of Neoclassical Growth Theory revised 2006 Background to the classroom presentation and discussion. © David M. Nowlan 2005 Unlike multiplier or export-base regional growth models, which are demand driven, neoclassical growth models are supply driven. Markets are assumed to work in neoclassical fashion so that production factors – capital and labour -- are kept more-orless fully employed; the source of growth in output therefore becomes the growth in the factors themselves, including a somewhat elusive “technical change” component which increases the productivity of some or all of the factors. Typically, two factors of production are considered, physical capital denoted by the symbol K , and labour denoted by L . A flow of labour input, say hours worked per year, works together with some stock of capital to produce a flow of output Y according to some functional relationship, Y F ( K , L) for some given time period. Leaving the technical change component aside for the moment, time subscripts can be put on the variables to remind us that the inputs and output are in fact functions of time. We then get Yt F ( K t , Lt ) Equation 1 The function F is quite general. This needs to be specialized in order to reflect the neoclassical view that factors of production are paid their marginal products and total payments to factors equals total output. So, in the neoclassical world Yt Y K t t Lt Yt or MPK K t MPL Lt Yt K t Lt Equation 2 In order for the function F to have the property shown in equation 2, it must exhibit constant returns to scale, according to Euler’s Theorem. This means that if you scale up each of the two factors of production by some constant proportion, say a , then output will change by the same proportion. If F (aK t , aLt ) aYt , then F exhibits constant returns to scale or CRS. A CRS function can have a number of forms but the most commonly used form is called a Cobb-Douglas function, which has the following specification: Yt K t L1t , where 1 Equation 3 You can check to see that if, for example, you double the input of both K and L , then output will also double. It turns out that the assumption of CRS, which is necessary within the neoclassical view of the world, and a Cobb-Douglas functional form has very important implications for growth theory. It entails the result that if one of the factors of production, say labour, is held constant while the other grows in size, the marginal product of the growing factor will gradually fall. More generally, if one factor grows relative to the other, its marginal product will fall. Thus the ability of additional units of the faster growing factor to increase regional or national output diminishes as the proportion of this factor increases. To see this, take the marginal product of capital derived from the Cobb-Douglas function, equation 3. (I’ve dropped the time subscript in equation 4 to make it easier to read, but it is implied.) Y L K 1 L1 K K 1 Equation 4 The equivalent equation for the marginal product of labour would be Y K 1 L L Equation 5 Notice first that the marginal products of both capital and labour are functions of the capital:labour ratio only; the absolute magnitudes are not relevant. This is a property of CRS functions. Notice further and importantly that, from equation 4, the marginal product of capital will fall as the labour:capital falls (i.e., as capital becomes more abundant relative to labour). From equation 5, we see that the marginal product of labour similarly falls as labour becomes relatively more abundant. Another way of looking at this is to note that with the Cobb-Douglas production function, Y output per worker, , which is often what we’re interested in as a measure of the well L 1 being of a region, is a diminishing function of the amount of capital per capita. To see this, take equation 3 and divide each side by the amount of labour: Y . But if it is assumed that the number of workers is a N L Y constant proportion of the total population, i.e., stays constant, then movements in will be exactly N L Y reflected in movements in . I will adopt that assumption and use growth in output per capita and N 1 Usually we look at output per capita, say growth in output per worker interchangeably. 2 Y K L1 K L L L Equation 6 Recalling that 1 , equation 6 shows us that a doubling of the amount of capital, for example, with the amount of labour held constant, would result in less than a doubling of output per worker. Growth in either factor alone is not enough to result in an equiproportionate growth in regional output. For example, the more you tried to increase capital by raising the rate of saving in a region, the less each additional unit of capital would contribute to a rise in output, as equation 4 demonstrated. More about this below. Notice from equation 6 that, with a CRS production function, we can write output per person as a function simply of capital per person; output per person depends only on the capital:labour ratio. Using lower case letters to denote per person amounts, we have from equation 6 y k Using equation 3, we can now look growth in this regional economy. Before doing this, a few points are worth establishing. First, notice that, with factors paid their marginal products, the share of each factor is equal to the exponent of that factor in the production function, i.e., the share of output going to capital is and the share going to labour is 1 . (Naturally the shares add up to one.)2 The rate at which a variable is changing at any point in time is given by the time dY derivative of the variable. For output, for example, this would be (or, if you don’t dt Y like derivatives you could write ; if this is the change for one time period, then t t would equal 1). The growth rate of the variable is this rate of change divided by the dY 1 level of the variable. Again for output, the growth rate of output is given by (or dt Y Y 1 ). t Y 2 To see this for capital, take the marginal product of capital, from equation 3: K L1 Y K 1 L1 K K Y . Multiply this marginal product by the number of units of K Y K Y K . capital and divide by total output to get capital’s share: K Y KY 3 For growth rates, I will use the symbol G . So GY denotes the rate of growth of output, GK the rate of growth of capital, G L the rate of growth of the labour force, G y the rate of growth of per capita output and so on. Finally, notice that the rate of growth of a variable may also be shown as the time dY 1 d ln Y derivative of the log of the variable. So, . (For proof see footnote.3) dt Y dt Back to looking at growth in this economy. The easiest way to do this is by taking logs of equation 3 and then taking time derivatives as discussed in the preceding paragraph. ln Yt ln K t 1 ln Lt Equation 7 From this, d ln Yt d ln K t d ln Lt , or 1 dt dt dt GY GK 1 GL 3 Equation 8 To show this set Yt Y0 e , where gt Y0 is some initial level of output, g is the growth rate of output and Yt is output at time t . Take the log of each side: ln Y t ln Y 0 gt ln e ln Y 0 gt , since ln e 1 . d ln Yt d ln Y0 dt Now take the time derivative: g 0 g.1 g . So the time derivative of the dt dt dt log is indeed the rate of growth. 1 A further note on the constant e used above. Recall that, by definition, e lim 1 , as . Now, consider some annual rate of growth, i , like perhaps the annual rate of interest on a savings account. t If you start with an amount A0 in the account then it will become At A0 1 i after t years. Suppose, however, that the interest rate is paid monthly and compounded. Then, after t years, i At A0 1 12 12t nt i . If the compounding is n times per year, we have At A0 1 . Let n nt it it 1 A0 1 . As the number of compounding times per year increases without limit, i.e. as n and so , this expression i 1 i 1 . Then we have At A0 1 A0 1 n n becomes At A0 e , a short-hand way of describing continuous growth at i per cent per year. it 4 In per capita terms, the growth relationship is this: G Y GY G L G K 1 G L G L G K G L Gk Equation 9 L Equation 8 says that the rate of growth of output in this economy is a weighted average of the rates of growth of capital and of labour. Remember 1 ; this means that the rate of output growth must always be between the rate of growth of capital and the rate of growth of labour, unless capital and labour are both growing at the same rate, in which case output will also grow at that rate. To keep things simple at this point, suppose that the rate of growth of population, and therefore of the labour force, in this region is given to us – it’s not something that we’ll derive from within the model. Let this given growth rate be G L . What about the growth of capital? If this is a region closed to outside capital flows, then any growth in the capital stock must come from savings in the region. Suppose the savings rate is a constant proportion of output, s . The annual addition to capital stock will therefore be sYt for any year t . The growth rate of capital is therefore: GK sY K Equation 10 Using equation 8 and these growth rates we can see how this regional economy will evolve over time. If capital is growing faster than labour, i.e., if GK GL , then output must be growing more slowly than capital, since its growth rate is always between the growth rates of the two factors of production, GY GK . This means that the right-handside in equation 10 must be getting smaller over time, i.e., the rate of growth of capital is diminishing. This will be true as long as GK GL . Since G L is given and fixed, GK as it declines must be converging on G L . Similarly, the rate of output growth must be falling, and it too will continue to fall until all three growth rates have become equal, i.e., there will be a convergence to GY GK GL . If we started from a situation where capital was growing more slowly than labour, then output must be growing faster than capital. From equation 10, this means that the growth of capital must be increasing. Once again it will continue to increase until all growth rates have converged, again to the labour growth rate. So, from anywhere we start in this model, the long term growth will always converge on the growth rate of labour. This may be illustrated with a diagram, Figure 1 shown below. 5 In this diagram, growth rates of the variables are plotted against the capital:output ratio, Y . The growth rate of labour, G L , is a given constant. The growth rate of capital, GK , is K sY given by . Where G L and GK cross, both growth rates are equal and this must also be K Y equal to the growth rate of output. Since is constant at this point, this can be K considered the long-run equilibrium level of growth in this region, shown as E in the diagram. Suppose the economy is not initially at equilibrium but starts at a lower-than-equilibrium ratio of output to capital, say at position 1 in Figure 1. At position 1, the labour force is growing faster than the capital stock. Output also will be growing faster than capital Y (although slower than the labour force), so the ratio will increase; the economy will K Y shift to the right along the axis until equilibrium is reached at E . K In the process of moving from the initial position 1 to equilibrium E , output per capita or Y Y Y per worker, , falls as rises.4 will gradually stabilize as E is approached. In the K L L Y long-run equilibrium, remains constant. L If the economy initially was at 2 in Figure 1, with a higher output:capital ratio and a Y Y lower output:labour ratio, then would, over time, fall to the equilibrium while K L increased. The growth rates in output per capita and the movement in per capita output itself, corresponding to the starting points 1 and 2, are shown in Figures 2 and 3. Suppose initial positions 1 and 2 represent the current situation in two otherwise identical regions. 4 In this model, Y Y and always move in opposite directions. We can show this. Starting from the K L Y L production function, equation 3, we get K K Y K L L L K and inversely with 1 1 Y K 1 1 L Y 1 which gives . Using this, we get K K K Y Y 1 K 1 . So, moves positively with L Y K Y Y . K 6 In region 1 output per capita will start off higher than in region 2. But notice that the higher output per capita in region 1 will be falling as the economy moves towards equilibrium. In poorer region 2, output per capita will be rising. Ultimately output per capita will reach the same long-run equilibrium level in each region. There will have been a convergence to this same level of output per capita.5 If this model approximates the real-world of economic regions, we should expect to see rich and poor regions, measured in terms of output per capita, converging towards each other – poor regions should be growing faster than rich regions. In the simplest case, the one we’ve been working with so far, this convergence should be towards a uniform output per capita. This is known as absolute convergence. We’ll discuss in class the evidence for and against absolute convergence across countries and across subnational regions. Sticking with our two regions, suppose that initially they are both in a growth equilibrium like position E in Figure 1. This is repeated in Figure 4. Now suppose that one of the regions, region 2 say, decides to try to improve its economic performance by encouraging, one way or another, a higher savings rate s in the region. If it is successful, s 1Y s 2Y the function showing the growth of capital in Figure 4 will shift from to . For K K this economy, the growth equilibrium will shift from E 1 to E 2 . Output per capita will initially rise but this growth will disappear as the new equilibrium is reached. At the new Y Y equilibrium, the ratio will be higher than initially, but the growth rate of will be L L just what it was at the beginning. (In this simple model without technical progress, the Y equilibrium growth rate in is zero. This equilibrium growth rate will become positive L when technical change is introduced.) Figure 5 shows the path of output per capita over time in these two regions. The two start Y off with the same level. After region 2 succeeds in raising the savings rate, its output L per capita rises but then reaches a new level with no further growth. Both regions have Y either remained at or converged to, once again, a constant rate of growth in (zero in L this case). Convergence to a constant rate of growth but at different levels of output per capita is called conditional convergence. In this case, different savings rates are the conditioning factor. Even in more sophisticated neoclassical models, this convergence to common growth rates remains the dominant feature. 5 This result is a consequence of the fact that with neoclassical production function, economies with a relatively low output per capita and correspondingly a relatively high output per unit of capital have a high marginal product of capital. Not only is the growth rate of capital relatively high in a region like region 2 (see the sY function in Figure 1), but the marginal product of each additional unit of capital is high. In the K other, initially richer region, the marginal product of capital is low. 7 Figure 1 10% sY K growth rates 8% E 6% GL 4% 2% 2 1 0% Y/K Figure 2 6% hi Y/K, lo Y/L 4% 2 Y/L growth rate 2% 0% -2% -4% 1 lo Y/K, hi Y/L -6% time Figure 3 10 (Y/L) 8 6 1 4 2 0 2 time 8 Figure 4 s 2Y K 10% growth rates 8% s 1Y K E2 GL 6% E1 4% 2% 0% Y/K Figure 5 10 (Y/L) 8 hi savings 6 lo savings 4 2 0 time 9 It’s time now to take technical change or productivity improvements into account. Based on recent economic experience we expect the productivity of factor inputs to improve over time. In many neoclassical models, this is captured by multiplying the Cobb-Douglas production function by a term, At , that grows over time: Yt At K t L1t Equation 11 In this formulation, technical change or productivity growth floats over the whole productive process embodied in the Cobb-Douglas function. There’s no indication of the origin of the technical change; in fact, the level and rate of change of At is given exogenously and is independent of such things as the amount of investment in capital. Because technical change is a multiplicative term in the production function, it can be associated with either the capital or the labour term. Because our interest is in the growth of output per capita, we associate it with the labour term in the following way: 1 t Yt At K t L 1 K t At1 Lt 1 K t Et 1 Equation 12 The term in square brackets in equation 12 is called the “efficiency units” of labour and, as you can see, these units are denoted by E t . As long as At is growing over time, the efficiency units of labour will grow faster than labour itself. Thus, 1 GE GL G A 1 The same reasoning that took us to growth equation 8, yields from equation 11 GY GA GK 1 GL Equation 136 This may be written as G GY G K 1 G L A G K 1 G E 1 Equation 14 G A is sometimes referred to as the growth in “total factor productivity” which, from equation 13, is GA GY GK 1 GL 6 10 The form of equation 14 is the same as that of equation 8, save that the growth in efficiency units, G E , has replaced the growth in labour units, G L . Now, equilibrium growth occurs where K and E are growing at the same rate, which will also be the equilibrium rate of growth of output. This time, however, this equilibrium growth rate of output will exceed the growth rate of labour, G L , as long as GA 0 . Output per capita will therefore be growing in equilibrium as a consequence of introducing technical G change. We have, in equilibrium, GY GE GL A . So the growth rate of output 1 GA capita, which is GY GL , is equal to . 1 The new situation, with technical change, is shown in Figure 6, which is similar to Figure 1. Equilibrium now occurs where the growth of capital equals the growth of efficiency units of labour. If two regions are alike in all relevant respects – the same savings rate, the same population growth and the same exposure to technical change, then the equilibrium of each will be at E in Figure 6. Even if the regions start at very different output: capital and output:labour ratios, say at positions 1 and 2 respectively in Figure 6, income per capita in each region will converge to a common level and a common growth rate. This again is an example of absolute convergence and is shown in Figure 7. A region starting off with a lower level of per capita income will grow faster than one with a higher level. They both converge to the same level and same growth rate. (Notice that in Figure 7, the vertical axis is the log of income per worker. A constant rate of growth therefore appears as a straight line.) Introduce once again a higher savings rate in one of the two otherwise identical regions. sY Imagine shifting in Figure 6 the line for one of the regions, as we did in Figure 4. K Y Y The region with the higher savings rate will end up with a lower ratio and higher K E Y and ratios, but the growth rate of output and of output per capita will, in equilibrium, L be the same in both regions. This is again an example of conditional convergence. The higher savings rate has introduced a level effect, i.e., the level of output per capita is higher in the higher savings region, but not a growth effect. Figure 8 illustrates conditional convergence with a level effect only. We saw in footnote 2 that the marginal product or return to capital is a direct function of Y the output:capital ratio, MPK . This leads to the insight that the more open are K regions, open that is to capital and labour flows, the more likely it is that absolute rather than conditional convergence will occur. Consider a region that has somehow achieved a 11 Figure 6 GK 10% E growth rates 8% sY K GE 6% 4% GL 2% 1 2 0% Y/K Figure 7 10 ln(Y/L) 8 starting at position 1 6 4 2 starting at position 2 0 time Figure 8 hi savings 10 ln(Y/L) 8 lo savings 6 4 2 0 time 12 higher rate of savings. As we have seen, this leads to a level effect in per capita income. Y It also leads to a level effect in capital intensity, the ratio: this ratio will fall as the K capital intensity rises, as you can see by looking at Figure 4 or considering footnote 4. This in turn means that the marginal product or return to capital in this region will fall relative to the region that didn’t raise its savings rate. Savers will now be attracted to the region with the higher return to capital – this is where they will want to invest – so some of the savings in the higher-savings region will be drawn into the other region. With complete openness to capital flows and other relevant circumstances identical, this flow of savings or capital from one region to the other will continue until the returns to capital sY are the same in the two regions. The adjusted function – the rate of growth of capital K – will then be the same in each region and the openness will have the result that the level difference in output per capita between the regions will have disappeared. Convergence will be absolute, not conditional, although because of the higher savings rate in one region, both regions will have a somewhat higher level of output per capita. Long-run growth rates will not have changed. Neoclassical growth theory thus predicts that output per capita of regional or national economies will converge over time. We should expect to see richer countries growing more slowly than poorer countries. While this convergence may be conditional on some factors that vary among regions, the more open the economies the more likely it is that convergence will be absolute. This theorizing is one of the principal bases for believing, in the European Union, that reduced trade and factor-movement barriers will lead to regional economic convergence. We will see in class that empirical support for absolute convergence among nations or regions is weak, but there is more support for the existence of conditional convergence. The most interesting recent writing on neoclassical growth has focused on the question of what factors exactly – higher levels of schooling, democratic government, rule of law, inflation rate and so on – are the conditioning factors that can produce level effects in output per capita. Notice that many of these factors are not transportable from one region to another through the open-economy mechanisms of capital and labour flows, so open economies alone may not be enough to bring about absolute convergence. We will explore some of these results in class. All neoclassical models predict convergence in growth rates. In spite of the considerable challenge to this position from models of economic geography and the new growth theory, the introduction of level effects and conditioning variables into the basic neoclassical model has allowed us to understand some of the factors that help a region become richer, or that might keep it poorer. 13