0,1 = 1, 2,3: in swhich family B ties to catch up to family A but fails to do s0, the sum of theseD, D, + miles = ri z traveled (wy ‘Stage 1 (aticr initial 60 minutes) a Dy les traveled miles traveled mH) CHOSH Dy Dy + + rites + traveled, 0 P, mR, (60) (90) (Ws) ct2.3) Stage 4 (after further t, = 7.5 minutes) * Dy 2D, Lu miles a . a traveled Stage i+ | (after further +, minutes) Figure 3.8 Relative location ofthe families in the Achilles Paradox story35 SERIES 95 Example 3.18 nite series which converges: Hea ele For the simple example above we know that the limit of this infinite series ts 460 (minutes), which is the time it takes for Family B to overtake family A.To see this formally, note that the series 1; has tems) = 30, to = 15.79 =7.5.-.-,.0F = ap! witha = 30and p= 0.5. Thus s; = Sot a iva geometri series and lion 5 Having resolved the Achilles Paradox by recognizing that an infinite sum of positive numbers may be finite should not lead one to claim oo much. Inthe geometric series, ¢, =a" with [p| < L-themth term approaches zero us > 20. However, it is ot generally correct that for any sequence ty With liMy-.:. the associated series s, 1; converges. ‘The series constricted from the sequence ay = 1/n, called the harmonic series, «does nol converge even though litt .2. dy = 0. (This is left as an exercise.) The intuition isthat if a seriesis toconverge, the nth term (a,) of tsussociated sequence ‘must approach zero “quickly enough’ that the sum of the terms ts finite even. as nn 6, This i the rationale of the condition stated in theorem 3.4, that ‘aya ‘which ensures that a series is convergent. Note that for the harmonic sequence Present Value of a Stream of Payments Earlier we saw how the sequence PV; = V/(1 +r’ (see equation 3.3) represents the present value of an amount of money V received ¢ periods into the future Im many economic setinss we need! (0 compute the equivalent present value of a seties (ie. the sum total) of such amounts, For example, a morgage or other long-term Joan represents & current sum of money loaned to an individual or96 CHAPTERS SEQUENCES, SERIES, AND LIMITS insttation in seturn for a stream of future payments. ‘Thus, {fan individual makes annual payments atthe end of exch year in amount V for 7 years, with the interest ‘ate being r, then the present value of this Stream of payments is e=y v See Oe NS rie ae Oe 36) “The relationships among the variablesin equation (3,6) te worked out in mort sage tables. They are generally computed on a monthly basis, in which case the appropriate interest rate isr/12, where isthe annual rate and T refers tothe num ber of monthly rather bun yearly payments. However, let us consider the following example based on annval payments, The yearly payment required to compensale {lender for a loan of amount $100,000 at an interest rate of 8% (» = 0.08) spread ver 25 years is V = $9,367.88, while if payments are spread over SO years the annual payments would te $8,174.28. Note that if the payments were to be made in perptity (ie, T —> 9c), we could Use the formal fran nfnte geometric sees with d= V/(L-+N. p = 1/(1-+rhsince Sy vito ov ity Fis Se a IN) 7 te = Dia Tiana 7 Thus we get V = rP = $8,000 for our example, which is Tasty close tothe valle when payments are made for 50 years. The reson fr this is iMustrated by the fact that the present value of $8.000 received 50 years from now, when the ieverest rate is 8¢, is omly $8.000/¢1 +008)" = $170.57, The present value of all payments of amknnt V = $8,000 received after $0 years ws tin So = Ji Py ~ tin ey A Dae which is only $2,132.13 in our example (100,000 — $97,867.87), «difference of only 2.1% (approximately), "Thus we see that using the formula for an infinite | series is often & good approximation for evaluating the present value of 3 finite series, provided that the number of periods or the interest rate isnot too small. The reason for the later requiresnent is that if i close to zero, the discount factor, = 1/(1 +9), is close to one, and so future payments are not discounted very ‘nvich and thus should wot be ignored. In fact, if the interest rate is only 1/2 of 1%. then we find that the present value of an infinite stream of payments of SK.000 per year is $1.6 million and the present value of the payments received afes the |Example 3.19 35 SERIES 97 ‘50th year is z Vv ie eae cad $1,246,858 which is 78% of the entire present value, compared to the error of 2.1% if the interest rate were 8% (rom the calculation above), Suppose that a stream of equal pay:nents of amount $10,000 per year is to continue in perpetuity. AL the interest rate of 6% compute (i) the present value of this entre stream of benefits (ii) the present value of che benefits beginning atthe end of the SOth year (ii) the present value of the first 30 years of benetits Solution oy $F Sica00 _ sto000 lim Pr mF (1.06 ~~ 0.06 $166,666.67 (ii) as of the beginning of the 50k year the present value of $10,000 per year in perpetuity is $166,666.67, as computed in part (). Since this isin effect eceived atthe end of the SOth year. its current (ie..8 of now) present value ‘must be discounted so that it becomes $166,666.67 (o6ys7 = $048.06 ‘The present value ofthe firs 50 years’ Worth of payments is simply the answer im i) Less that in Ci $3166 666.67 ~ $9,048.06 = $157,618.61 . “The examples above all dea! with the problem of determining the present value ‘of @ series of equal payments. In general, however, one can evaluate the present value of any pattern of payments. Suppose, for example, that a business firm is ‘considering the possibility of making acurrent (and immediate) investment of $C, the payatf of which will be the sules revenue af a product whose sales will increase lover time. Let us assume that the production process will begin at the end of one98 CHAPTERS SEQUENCES, SERIES, AND LIMITS. Example 3.20 {year and that net profit from sales of the product is Sr(1 + g) the first year and ‘will grow at a rate of g each subsequent year. Thus the profit for period ¢ will bea, = 711 +g, and the (undiscounted) valve of the stream of benefits (gross benefits) will be ‘which isa divengent series if and g are positive. The discounted or present value ‘of the stream of benefits is 3a) which is just geometric series, with @=(h +g)/(+r) and p=(14+8)/ (+2), $0 that PVB = m(l + g)/(r ~ g) and is finite valued if and only if C or PVB- ere Lease PY of costs PY of revenues ‘The tight side of this equation represents the amount $70,000 every year in per petuity beginning five years from now. As of the beginning of year 5, this has 4 present value of $72,000/r, and so discounting this number to account fora Five-Year delay before it begins means that it has a current value of (870,000/r) ae } Tis not easy to solve analytically for i this refaionship, which can be written $50,000 __($70,000/r) $100,000 + r Cr PV of costs PY of benefits By using a computer to ty several values for, we find the solution to be r = 0.05 Values less shave r = 0.05 give a higher present value for benefits than costs, while ‘values greater than r = 0.08 give a higher present value for costs than for benefits ‘The reason for this is that the benefits of this project accrue relatively further i the future and so are reduced to a greater extent than are costs as the interest rate rises. Table 3.1 illustrates this poim, Table 3.1 Interest rate (¢) PY of benefits 0.04 $1,350,000 $1,438,000 0.05 $1,100,000 $1,100,000 0.06 $933,300 $871,80035 SERIES 103 ‘The Keynesian Multiplier An important component of the traditional Keynesian macroeconomic model used 1 explain the importance of government fiscal policy isthe multiplier. The basic idea is that if there is an increase in exogenous expenditure in the economy say Bovernment spending, then # multiplier effect ensues so that the ultimate impact ‘on economic activity (GNP) is greater than the initial expenditure. For example, suppose that the government initiates additional expenditure of $100 million for increased road repair. This expenditure then becomes additional income for fms and houscholds. Although we would expect these individuals to save a certain fraction of this extra income, pay part of it as taxes, ane perhaps use some to purchase imported goods, we would nonetheless also expect that a certain fraction \wouild be subsequently spent on domestically produced goods and services. Ifthe fraction of this income spent domestically is 60%, then the increased goverment expenditure will have a second-round impact on domestic incomes of $60 million {ie.. $100 million 0.6). The individuals receiving this additional income of ‘$60 million will also spend a certain fraction domestically. If hi also 60%, then the third-round effect is equal 10 $36 million ie, $60 million x0.6), Continuation of this provess is described in table 3.2: Table 3.2 Ist round effect (initial) 100,00 million 2nd round effect 60.00 million (100. 0.6) 3rd round effect 36,00 million (60 « 0.6) 4th round effect 21.60 million (86 x06) Sth round effect 12.96 million (21.6 x 0.6) and soon Wecan seein the table that the overall impact can be described asthe geometic series 100 + 1010.6) + 100(0.6}? + 100(0.6)° + += which, from equation (3.5), has the value (a = 100, » = 0.6) 00 _ 100 06 04 250 million ‘The fact that even if we allow for an infinite number of rourids in this problem, the overall increase in national income is finite because of the same intition that resolves the Achilles Paradox. Summing an infinite numberof terms, each positive. imay give rise toa fini value ifthe rerms become small sufficiently fast.Example 3.24 Suppose that there isan initial increase in government spending of $100 billion tnd (hat individuals spend 80% of any extra income on domestically produced goods. Use the Keynesian multiplier model to determine the overall impact ofthis $100 billion injected inte the economy. Also compute the fist five rounds of the process as was done in table 3.2. Solution ‘The first five rounds of effects are described in table 3.3. The overall impact is 100 100 + 1000.8) + 100(0.8)° + 100(0.8)" + 3 = 5M" billion Table 3.3, {st round effect (initial) 100.00 billion 2nd round elfect 80,00 billion (100 x 0.8) 3rd round effect 64.00 billion (80 x08) 4th round effect 51.20 billion (64 08) ‘Sth round effect 4096 billion (51.2 « 0.8) In general, if we let c be the feaction of extra income that each individual \would spend on domesticaliy produced goods (0 ° () me G10) Which tums out to be a series that converges (1e., has a finite-valued limit). We ‘can see this immediately by using theorem 3,4, Thetis, = tg Leno a Ten tim 2 tet In Ln Sa wy lim‘which is fess than J, and so the series represented by equation (3.10) converges, (Note that the second fine follows from In(2"*")=(n-+1)In2 and In(2") ‘min2.) Thus, by using the function In(y) 10 represent the utility of $y and pre- suming that individuals rank risky alternatives by comparing the expected value of the utility they generate, rather than the expected monetary value, we avoid the problem of concluding thatthe St. Petersburg gamble is worth more than any certain amount of money. no matter how large in value. ‘Examining this issue further is worthwhile as itbrings out some ofthe intuition bout infinite series involving terms that tend to zero or infinity. Let pg be the sequence that represents the probability that the first head occurs after 1108508 G.e., py =1/2") and let ey = (5) be the utility Bain upon receiving the winnings Of ip (i Yq =2" and w(-) = In(-) forthe example above, Thus represents she expected utility of the gamble. The limit ts the sum of the product of the terms from two sequences, one that tends fo zero (p,)-and another that tends {0 infinity (up) a8 n —> 00. The series converges if the ui. ferms do not approach infinity 200 quickly relative to howe quickly the py terms approach zero. We saw above that if Ye = 2" and we use iy = Yq (Or any linear. increasing function of y,, then the series does not converge, while if we use x. = n(y,), the series does converge. The problem with this resolution of the St. Petershurg Paradox is that fone can construct a sequence of prizes. such 8s jy =e”, such that the series will no longer converge even if u(+) = In(:). This follows since i which isa divergent series (recall that In¢e) = 1). Thus. by altering the prize sinu ture, we see that adopting the utility function u(y) = Inty) will nol always resolve the type of paradox raised by the St. Petersburg gamble. A suflicient condition for resolving this paradox is to assume that the utility {function is bounded, f(y) is bounded above by the value u* as illustrated ir108 CHAPTERS SEQUENCES, SERIES, AND LIMITS wy) ou is) Figure 3.10 A beundod wilty Function figure 3.10, then the series Eu am ‘An alternative way out of the St, Petersburg Paradox is to assume that the prize values are bounded, say because of a budget constraint. aaa 1, Consider the trivial sequence a, = c, ¢> 0.8 constant. Show that this is aa example of a sequence for which and the ries generated by it diverges (see theorem 3.4), 2. Show that the harmonic series diverges. (Hint: Group terms from i = 2 +1102", k = 12,3... aand note that the sum in each group is greater than 1/2.] Show that this ix an example of a sequence for which lity last /in] = 1 and the series | generated by it diverges. By writing out and expanding expressions for sy and p’ ll 0, however smalt, there exists some 5 > 0, such that | fic) —L#| 1/6) tn fact, ifthe left-hand limitis L", then the sequen of values f= (4s) must converge (0 the value L for, =u ~ (I/n) ‘This example of «sequence of values xy that approaches. x =a from the left illustrates the relationship between limits of functions and limits of sequences (see chapter 3) and also explains why the notation x > a is used to denote a left-hand limit (x =a minus some small positive value). One must realize, however, tha {or the left-hand limit ofthe function to exist and equal L, jtmust be the case ‘any Sequence of ty values approaching a from the left must induce a sequence o function-ealues fy = fq), which converges to the limit L!. ‘A similar definition applies for the right-hand limit, namely that if there i always some (possibly very small) range of values of x to the right of a (iss strictly greater than a) for which the function values f(x) can be made ar trary close to some value L®, then we say that the right-hand limit of (1) a x=u exists and is equal 10". An example of a sequence of x-values that proaches.x = a fromthe rightis.xy=a-+1/n. n= 1.2, 3,.... Hence we have rotation lin, Fes‘The right-hand limit of a function f(x) which is defined to the right of x the point x =a, exists and is equal to L®, written lim, f(x) = 1" iffor any € > 0, however small, there exists some 5 > 0, such that |/(c) — L*) < e.¥ x, satisfying a 0 such that |f(x) — /(a)| <« whenever |x ~ a <4 forany € > 0, A funetion that is not continuous is said to be discontinuous. Definition 4.3 is the more useful definition when trying to show that a function is discontinuous at ‘acertain point. This will become clear when discussing why the examples of the118 CHAPTER 4 CONTINUITY OF FUNCTIONS: Example 4.2 fa) 646, 6 6-4 383335 ¢ Figure 4.1 Function y ~ 2x (with demonstration hit iis Continuous at r=3) Example 4.3 functions illustrated in figures 4.3 104.6 are discontinuous. Each case demonstrates «distinct property that i the cause of the discontinuity. Together the cases exhaust the reasons why a funtion miay be discontinuous. Itis often simpler, however, to use definition 4.4 to show that a function is continuous, as will be seen forthe examples of continuous functions illustrated in figures 4.1 and 4.2, Definition 4.4 illustrates the similarity between the definition ‘of continuity of a function and that of the limit of a sequence (definition 3.2), Ac cording to detinition 4.4, a fmetion f(x) is continaous at the point x = a if (1) is arbitrarily close 10 f(a) forall values of x close fo x =a (ie.,for \x—a) <8) While a sequence f(r) has limit value Lif f(n) is arbitrarity close ro 1. forall values n heyond some value that is sufficiently large (ie.. > Show that the Tineas function f(c) = 2c is continuous, Solution ‘Take the function f(x) = 2x. This function is continuous at every point x & R, To see this, consider the point x = 3 in figure 4,1. The function value at x is (3) = 6. According to definition 4.4, this function is continuous at this poiat if, no matter how small a number e > 0 we choose, there is some value 8 > 0 (possibly very small) such that all the function values defined on the set of x values (3 ~6, 3 +8) lie within the set 6 ~¢, 6+) Intuitively this means that the function values defined at points near.x = 3 are all clase to /(3). For example if we choose € = WAL then by choosing = 0.002, we find that all funetica values f(x) defined on the set x & (3 — 0.002, 3 + 0.002) lie within distanae € of f(3). That is. fx) © (PG — 0.002), 3 + 0.002)), which implies F(x) € (6 — 0.04, 6 + 0,004). All of these values are within distance € = 0.0 if (3) = 6. This would be true for any choice of § < 0.005. More generay. see that at any valve « = a we can be assured that / (x) will lie within distances ‘of f(a) as tong as we choose 3 < €/2. Thatis for s € (a ~ €/2,a+€/2),We fly € Qa—e,2a +6), Infortwally then, a funetion is continuous at a point x = a if the function defined at that point (i. F(a) exists) and you can meet the following challenge. matter how smal value € > Dis chosen. its always possible to find suficie ‘small but positive 6 such that forall the x-values -ithin distance 4 of the pois x =a, the function values fall within distance € of f(a), ‘Show that the polynomial function (x) =x? ~ 2x? +1 is continuous atthe x=1/2.Example 4.4 Jia) Ta=reo “isso Figure 4.3 Function in twample 4.4 (which is discontinuous ur=0) 4:1 CONTINUITY OF A FUNCTION OF ONE VARIABLE 119 Solution ‘The graph for the polynomial function f(x) =x*—2x?+ 1 is drawn in figure 4.2. At c= 1/2, we see that the function takes on the value. f(1/2) = 5/8, tis clear from the diagram that if for any € > O wean find a value 8 such that for xe (1/2— 5, 1/2-+8), the function values will ie within the set f(x) € (5/8 —€.5/8 +) In this case the actual choice for for any given ¢ is not s0 easy to ascertain, Some reasonably straightforward cases are given in the exercises following this section Figure 42 Function = 244 | (with demonstration that its continuous at r= 172) . Show that the function below is discontinuous: soo = Solution ‘This function is graphed in figure 43. Notice that the function has an obvious “break” or “jump” at the point x = 0. AS one approaches the point x = 0 from the left, the value of the function is +1, while as one approaches x. = 0 from the right, the value of the function is —1. In other words, the left-hand timit, lim, .o- f(2) = +1y is not equal to the right-hand limit, fim, .o- f(x) = —L. ‘Thus condition (i) of definition 4.3 is not satisfied atthe point x = 0, and so the function is not continuous at « = 0. ‘We can also use definition 4.4 to see that this function is discontinuous at the point x = 0. Suppose that we try to find an interval (open set) of x-values including the point x = 0 such that for all x-values in this interval, the function value is within distance ¢ = 0.5 of (0), We will fail, No matter how small a value 8 > 0 ‘we choose, we will aways find that f(r) will take on both the values +1 and —1 for some x € (—8, 8). .Before considering further examples of functions which are discontinuous, its useful to introduce the concept of the asymptote of # function Ifthe value of a function /(x)..x € R, becomes unbounded! as x approaches some ‘value x=a either from the left or the right, then we say the line x = a is any asymptote of the function, In such cases the funetion is nor continuous. If any one or more of the following possibilities holds, we say that the line x = q isa vertical asymptote, lim, f@) = +00, im fod Jim for) = 00. of tim fee 1 /x? has the same left-hand and right: = 0, itis not continuous, Example 4.5 Show that although the function fx; hhand asymptotic limits atthe point Solution fay ‘The line x = Gis 2 vertical asymptote for the function f(x) = 1/4 because the left-hand and right-hand limits of f (x) at x = © are bath +90: fay =e Jim fe) = Jim f(x) = 48 ‘We say that the asymptotic limit ofthe funetion-at x = (is +20, This possibiiy is similar to the idea of @ sequence being definitely divergent, as discussed in chapter 3. Although the left- and right-hand limits of the function are equal, the fanction is not continuous because f(@) is not defined. Thus condition (i) Figure 4.4. The function definition 4.3 is violated. The same holds true for definition 4.4 which seq) F(x) = Ve" Oohich is discontionsns that (Gr) must be defined on some open interval including the point x = 0 fori atx =Osinee 1/1 isundefined there) to be continuous at x = 0 (see figure 4.4) Example 4.8 Findanexample ofa functionthathas diferent eft-hand and right-hand asymptote limits at a point and so is not continuous a that point Solution ‘The function f(x) = |/(x — 1) has different left-hand and right-hand asym limitsatthe point. x = 1 Fors sghtlylessthan 1. itfollowsthat x ~ 1 isasmall ative number, and se L/(x ~ 1) isnegative but large in absolute value. For x sighExample 4.7 fis) =3x.982 : 322 | Fgure 4.6 Function in example AF (which is discontinuous at x = 2) cater than 1. it follows that (x ~ 1) isa small postive number, and so 1/(x ~1) isa large positive number, Thus lim,...\- f(x) = —oo, while lim, .)- f(x) = +00, asillustited in figure 4.5. As was the case for example 45, tis Fonction is not de fined at the point in question (x = 1) and so is not continuous at. = 1. In this case there isthe additional problera that the left-and right-hand lint are not the same. soa fo) = Merb Figure 4.5 Function fx) = /(x — L) (which is discontinuous atx = since 1/1) is undefined there) 2 Show that a function with a hole init is not continuous. Solution Consider the function ax, WreRa #2 no=ts” ¢ ‘which is graphed in figure 4.6. Despite the fact that this function is defined at (2) and its left- and right-hand fimits are equal atx = 2, the function is not continuous here because condition (ii) of definition 4.3 is not satisfied ‘We can also see formally that this function is not continuous by using defi- nition 4.4. At the point x = 2 we have f(x) = 3. However. consider the value Itis not possible to find an interval of x-values, (2 — 8,2 + 8), for which the function values are within distance € of /(2) for every x in this interval, This is true no matter how small a value for 8(8 > 0) is chasen, . ‘The following results, which summarize some useful properties of continuous functions, ae intuitively straightforward. We provide a proof for only one of them, which should give the Havor of the rest ofthe results,Proof (part @) only) Since itis assumed that /(2) is continuous, we know from, nition 4.3 that for any x = a. limp F(x) exists and is equal to ja). That Himy a> f(x) = Vimy o- F(4) = Fla). Thus, letting Aix) = cf Cx), we get =cfla) lim Ce) = tim ef (e) =e tim fe Similarly tim, 9: A(x) = ef (a) and h(a) = ef (a). Thus al ofthe conditions continity are satisfied forthe function Mx) = ef U9. ‘There are many instances jn economies in which the domain of a function ‘not the entre real line but isa proper subset of it For example, in considering amount of labor Z. fora fim to employ. it does not make sense to consider neg ‘values. Thus the relevant domain of a production function g = Q(Z) inv a lower bound or boundary, and we would write L > 0 or L € [0,ce). ‘considering continuity of the function Q(L) atthe point L = 0, it does aot sense to think about an open interval containing that point. In other instances domain may have an upper bound or both a lower and upper bound. For exampl in the short run a firm is constrained by its existing capital stock, say K < Ky so the relevant set of values for capital to choose from would be K€ (0, R]. ‘To deal with eases where the domain has boundary points, we make use the concepts of right-hand and lefi-hand limits as giver in definitions 4.1 4.2. Thus suppose that the function f(x) is defined enly om the closed in xa, b], a a~, and since x cannot be greater than b, we cannot def the right-hand limit of f(c) as.x —> b*. However, as long as (x) is continu al every point strictly within the interval [a.6], meaning for w a (from the right) andr + b~ (from the let), we say the fuaction (x) is continuous on the closed interval [a,bLet f(x) be defined on the closed interval [ 6) £2) is continuous from the right at the point x fa) exists, and lim, 9s f(x) = Fla, (ii) (2) is continuous from the left at the point x = 6 if lim, .y- f(x) exists, F(b) exists, and Lim, i> F189 = £16). Gli) C4) Is continuous on the closed interval [a,b] if it is continuous at every point x strictly within the interval (i.e., 0, x € R)? In the frst place, ro assume that f(x) is continuous ‘ata point x = a implies that J (x) is defined on some open interval of real numbers containing the point u. This means that must be infinitely divisible. That is, one ccan choose x to he a value that deviates even by infinitesimal amounts from ¥ =a. ‘An example of an input (and in output) that would not be infinitely divisible ‘would be bolts used in the production of an automobile. Since one would net use a faction like a half of a bolt. it Would only make literal sense to treat bots ay integer valued, Therefore it does not make sense to conterapate an open interval ‘of points including some value x = « bolts. However, if « manufacturer produces 20.000 vehicles per year using 1,050 hols in each vehicle, it seems reasonable to Simply treat bolts and vehicles as infinitely divisible and represent the relationship hetween them as y = /1,050, where.» is the number of bolts used and y is the number of vehicles produced. (Of course, we have ignored all he other inputs.)126 CHAPTER CONTINUITY OF FUNCTIONS Figure 4.7 illustrates this function white the true relationship would inchide ‘only the points (1,050, 1), (2,100, 2) (3.150, 3), ete. IF one ses the continuous funetion y = (1/1.050)x in the proves of solving some decision problem forthe firm and discovers the solution invalves some value of tht is not a multiple of f 1,050, then using the elosest value that is « multiple of 1,050 would probably he i reasonably uacurate. Thus, even if a commodity is not infinitely divisible, we may i often assume that i is, witht «storting reality very much, 1,080 2,100 31507 A Salary Schedule with a Bonus Payment Figure 4.7 Ye Cunction ‘Suppose that a salesperson receives a salary according toa contract that establishes y= (1/1.080 4 celationship between pay and the level of sales nade by the salesperson. Ia particular, suppose thatthe contrat stipalates thatthe salesperson's monthly salary will be composed of three parts: (i) abasic amount of S800, (ii) a commission ol 10%, and (if) & lump-sum bonus of $50) if the salespersons sales for the month reach or exceed $20,000, From this description, one can see that her salary will jump by $500 if the critical level of $20,000 worth of sales is achieved. This implies a discontinuity in her salary schedule, Letting $ represent sales per month and P cepresent the salesperson’s pay for the month. it follows that the function) describing her salazy-saes celationship is $8000.15, 5 = $20.000 $1,300-40.18, § > $20,000 Which is ilustrated in figure: 4.8 - 3,200 2,400 1.000 P= 400+ 0.15; s< 20.000 20,000 Figure 4.8 A salary schedule with s bonus paymentFigure 4.9 A discontinuous jnsme support function ‘The fact that the bonus of $500 is achieved once S reaches the critical value ‘of $20,000 leads to the result that the left-hand limit of the salary function at 5 = $20,000 is $2,800 while the right-hand limit is $3,300. The existence of this discontinuity has interesting economic implications. Consider the following scenario, There are three salespersons, called A, B, and C. Their cumulative sales for the month, not inclusing the last day, are $26,000 for A, $18,500 for B, and 36,000 for C. The 10% commission on sales will give each a similar incemive to inake extra sales onthe last day af the month, But will the $500 bonus possibility have a different effect on the three salespersons? Assuming that it is plausible Renerate a few thousand dollars worth of sales in a day but vinually impossible to Cereate more than $10,000 worth of sales. one woul expect that salesperson B will try harder onthe last day to increase sales than will the other 10. A Discontinuous Income Support Program Many welfare programs or “income support programs” offer individuals who are not employed a fixed or lump-sum monthly payment that is made only if the individual does not earn any income, Once ar individu earns any income whal- soever, the payments stopped. Corser te following hypothetical example. Sup- pose a single parent of zwo preschool-sged children can collect a monthly welfare ‘payment of $750 provided she doesn't work. However, once she earns any positive miount of income, the welfare payment stops. Assume that she could earn $15 per hhour at some job for which the number of hours Worked per month is entirely flexi- ble. The income ofthis person, asa function of hours worked, i. is given below: vay = [7 h=0 = Tish, hoo ‘The graph depicting this person'sincome asa function of hours workedis provided in figure 4.9. Its clearly discontinuous at /i = 0 hours worked. ‘Thistype of discontiauity, which isa property of many “all ornothing” income support programs, has been he subject of a great deal of debate. One ean see that a ‘person in sucha program would have to work 50 hours per month just vo match the income earned from the support payments. Since the person Would face childcare and other costs of working, the “all or nothing” property of this program presents a serious deterreot to the incentive to work. An altemative scheme would be to allow a person in this situation to keep) 4 certain fraction of income earned in addition to the $750 monthly payment Suppose. for example, thatthe person were allowed to retain 530% of any earnings. with the other 50% representing a payback of the income Support up fo the level ‘where the entire $750 is paid back. A person facing & Wage rate of $15 per hour ‘ill have paid back the full $750 only after working 100 hours oF more per month (0.5 x 100 x 15= 750). For 0 << 100, the effective wage rate is 50% of $15 (0F $7.50), and so net income for this range of hours worked is ¥ (1) = 7.5h. AfterCHAPTER 4 _ CONTINUITY OF FUNCTIONS this amount of earnings, the individual would keep any excess. Therefore undet this program the person’s income schedule would be the following: 750+75h, 0 100 ‘The graph for this income schedule is provided in figure 4.10, Notice that it is continuous. [Check that limp. ¥(h) = limy..,oy- ¥(h) = ¥(100).) vu) ry (2s 7s a0 15h h> 100, 750 100 D Figure 4.10 continnoys income-sipport function | Many economists prefer this second plan because it avoids the discontinuityl) the first plan. tn the first plan there is effectively large penalty for working a il since income drops from $750 per manth to almost zero ifthe individual ch ‘only a few hours of work. Under the second plan the person always earns moe) income by choosing work more. The resus that he person will be more lke choose some positive hours of work under the second plan making himsellhersl better off and slso reducing the cost ofthe program (o the governmet Continuous Marginal-Product Functions ‘The marginal product of an input is Ue amount by Which oxtput increases a result of an addtional uoit of that input being used, given fixed amounts ‘other inputs available. This concept is useful in economies when analyzing Accision-making problem of firms in the short run when the level of some in can be altered (variable inputs) but sther input levels are fixed (xed inputAull reatment of the marginal product of an input will be taken up inthe following chapter on derivatives. However, there are some interesting problems concerning the continuity of marginal-product funetions that are useful to consider here. For example, suppose thatthe function y = 10L-relatesthe anhount of output produced, +, the amount of labor input employed, L, for given fixed levels of other inputs. ‘ne can then see that an increase in L of one unit always leads to an increase in output of 10 units. Thus the marginal product of labor function is the constant function y = 10 and so is continaous on the interval (0, 20) Notice that this marginal-product function has the rather unrealistic property tha more output is generated by using more laber even for very large values oF labor, Since the amounts of all other inputs are fixed, one might anticipate that ‘makes more sense to imagine that as Z inczeases, the added output generated begins to fall and may even become zer0 oF negative. The following discussion shows that this phenomenon may occur in such a way tha the marginal-product function should be modeled as adi Marginal-Product Function with a Capacity Constraint In many production processes the maximum output that can be produced by in- creasing the amount of a variable input depends on the amount ofthe other (xed) inputs available. A good example is a coal-fired electricity generating station. ‘There will always be some absolute maximum amount of power that can be gen- crated from a single station. This maximum is generally referred to as the expac- ity of the station. For example, if station has & 1,500-megawatt capacity, 1 means that no matter how much coal or other inputs ate ayailable, the maximum. amount of energy that can be generated in a twenty-four hour period (per day) is 36,000 megawati-hours (i., 24 x 1,500) Suppose chat we want to determine the marginal product of coal for a ease in which there is enough of all inputs other than coal to keep a 1.500-megawatt power plant operating at capacity. Assume that it takes 250 pounds of coal to zeneratz one megawatt-hour of energy. and so one ton of coal will generate eight megavatt-hoursof energy. Therefore the marginal product (per day’ of coals eight megawatt-hours (per ton) as long as capacity has not been reached, Once capacity has been reached, however, the marginal product of coal draps to zero, Thus, once 4,500 tons of coal have been used in a day to generate electricity, the generating station will have reached capacity (4.500 x 8 = 36,000), If we let. represent ions of coal used per day and y the manginal product of coal in megawatt-hours, then the marginal product of coal is given by the funetion 8, 051 =4.500 0. 4.500_y- meguatt-hours of electricity per ton of coal 8, 05554500 0, ¥>4.500 4500 ‘tons of 608 Figure 4.11 Marginal product of a input under a capacity constraint Revenue Function, Cost Function, and Profit Function for a Perfectly ‘Competitive Firm In the model of perfect competition it is assumed that each firm treats the market price as given. The firm does not believe its own choice of output level will influence the market price, and soit treats this value es fixed. This assumption is usually only ‘made to describe markets in which a large number of producers each produces & small amount of some homogeneous (identical) product. Letting 7 be the price of the product and y be the firm's output evel, we gt the revenue function R(y) = Py. Since this isa linear function in y. itis continuous. To show what must be tne in order that the cost function, C(v), be continous takes i lite effort. We begin ‘with the long-run situation. ‘The total cost of producing a given output level is simply the cost of all the inputs used in the production process. Suppose that there is only one input x wed to produce y and that the production function y = f(x) is used to describe the production relation. If we know how much output can be created from variows levels of input, then we can work backward to determine the level of input needed to produce a given level of output. This simply generates the inverse function y=fu) = x= sf as demonstrated in igure 412 for the ease y = x13, “Thus the cost of producing output level y is wie, where x is the amount of input required to produce output level and w is the persunit price ofthat inpyesay ‘ip msi Figure 4.12 A production function and is inverse Using the inverse function x = f-"y), we ean wrile Cl) = wf!(y). Now, if the production function f(s) is continuous, then, because of theorem 4.1(¥i), s0 is its inverse, f(y), and $0 €(y) is continuous. The profit function, (1) R(y)—C(y)isalso comtinusus by theorem 4, (ii, This is illustrated in figure 4.13, for a more general cost function. 12G)). RU CO) THO) = Ri) —Cy) Figute 4.13 Ifthe cost Function. C(y) is continuous, so js the profit function.) ‘i the short-run situation we usually model the firm’s decision in terms of a single (Variable) input. Again, let represent this imput and y= f() the (short- fun) production function, ‘The same analysis as in the above paragraph applies except that we need to recognize that if the firm produces. no output ia the short ‘un, itstill must pay its fixed costs. If we let fixed cost be SCy, then the cost function becomes C(,)) = wf “"y+Cp. Once again i follows that if f(x) isa continuous unetion, then so is the cost function and then so is the profit function. An example is illustrated in figure 4.14,12 CHAPTERS CONTINUITY OF FUNCTIONS Fy), RO}, CO) a Figure 4.14. cos function and profit function inthe presence of xed erst ‘Consider the following case in which the profit function is not contimuoes ‘Suppose that a firm cannot change from zero preduction to pesitive production ‘without expending resources to start he production process. Forexample, a smiet- ing furnace may have 16 he preheated before any steel can be produced. This type ‘of cost is called a serup cost and differs from the usual som of east in that it isa Jump-sum cost that must be incurred When going from zero to any positive ammount cof production, The amount prexiuced does not affect the size of this cost, The result 's thas the cost function, and hence the profit function, will be discontinuous at 1 = 0, Letting c(y) represent the cost of producing y. excluding the setup costs, ‘we cn write the cost function as 0 a Bey). y>0 cu) Where By is the setup cost. Then the profit function is, ayy = {° y=o TS | RO) = Bo et), yO ‘These Functions are illustrated in figure 4.15. {comparing the cost functions in figure 4.14 with that in figure 4.15, itis important to distinguish between fixed costs, a short-run phenomenon nil and ‘elup costs. The existence of fixed costs does not lead to # discontinuity of the short-run cost function because. if the firm chooses to produce zero output, itisCO). ROTA oH Figure 4.15 A cost function and profi function in the presence of setup vats still the case that the firm must pay its fixed costs, Thus, for the cost function represented in figure 4.14, che right-hand limit of €(y) at y = Ois equal to Co; that {sto say, limy_g+ Cy) = Co, which is alsa equal to the value of the cost function at y = 0. For the cost function illustrated in figure 4.15, the setup costs, By, are incurred for any level of output no matter how small or large but are avoided when the firm chooses to produce zero output. Thus, the right-hand limit of C(p) al y = D for this cost function is equal to By, namely lim,-.o» C(y) = Boy but the value of the cost function at y = 0 is 0, Clearly dhis function is discontinuous 2: the point y = 0, ‘The Bertrand Model of Price Competition* I there is more than one produceriseller ina matket but nt So many as tor make the perfectly competitive model applicable, we say that the market structure is ‘oligopolistic. The word oligopoly means “few sellers.” One model that describes the possible behavior of firms in this seting is the so-called Bertrand! model. To make matiers Simple, we assume that there are two firms in the market, although the qualitative nature of he outcome of the model is not altered if We assume that there are more thaa two firms, Tn the Bertrand mode! the twa firms ure assumed (o compete in prices. That ceack firm sets price and then meets whatever demand exists for its preduct a that price, Assuming thatthe firms produce identical commodities, fone firm charges. lower pre than the other, then all the consumers will purchase from that producer. Ifthe two firms charge the same price, then we assume that consumers194 CHAPTER 4 CONTINUITY OF FUNCTIONS purchases will be split evenly between the two producers. Thus we need to about how revenue for each firm changes as prices are altered, To see how firms will behave in this situation, consider the following simple numerical example: Lot the demand function be y = 20 — 2p, and let the marginal cost (i, the cost ‘of producing one more unit of output) be the constant value ¢ = 4 for each firm, ‘That is, Cy) = 4y is the cost function for each firm. Nowe that as long as the price exeeeds 4, each firm cant make an excess profit. If the price equals 4, each firm can only earn normal (or zero economic) profit, and ifthe price falls below &, either firm would incura loss iit produces any output. Therefore each firm would Produce zero output, We begin the analysis by determining how firm ’s revenue changes for alter: native prices given that a specific price has been set by firm %, Sty p> =7. Given that firm 2 is charging the price p2 = 7, it follows that if firm 1 charges a price “above 7, its sales will be zero and so will its revenue. If ic charges a price equal 07, it will share the market with fiem 2, To find total market demand, note that at a market price of 7 we have 20-217) =6 Since the wo firms share the market equally when they charge the Same price, we have yi mn 40 firm I'S revenue is RCP = iyi = 78) = 21 implying a profit level of ay = Rp) ~ Cul) = pss —4y1 = 78) — 43) =9 As noted, if firm 4 charges 2 price even slightly above 7. it loses alla its market hate to firm 2 and so its revenue and profit drop to 0. IF firm I charges a price even slighily less than 7, however, firm 1 will capture the entire market and so its revenue will jump accordingly. To see this, let p, = 7 —e, with « positive bat small, Since p is less than the price changed by the other firm, firm 1’s sales will be determined by the total market demand. Thus firm I sells output level ye =20— 2p, = 20-27) = 6 $2 sand so earns revenue Ri = Pry = (7 ~ EMG + 2c) = 42 + Re De? sand profit 4) —4(6 426) = 18-2 42 + Be m= RUPEECONOMIC APPLICATIONS OF CONTINUOUS AND DISCONTINUOUS FUNCTIONS 135 For e small (ie.,¢ — 0), We find that firm | earns revenue Ry m= 18 In fact, forany price p; <7, firm L captures the entire market and soits revenue and profit functions become that of a simple monopolist. We can therefore weite firm 1's revenue funetion as 12 and that profit 7(20—2py, pr <7 Rim = 421, m “aby 0 pot Firm 1°s profit funetion is i202) —A20—2p)). py <7 mip = {9 (ay 0, The revenue and profit functions are illustrated in figure 4.16, Looking at the revenue function and returning to the conditions for continuity given in defini tion 4.3, we see that this is a particularly interesting case mathematically. At the Point 7) = 7. the left-hand limitof the revenuie function is42, the right-hand limitis D.and the valve of the function itself is21. Thus we see that lim, .7- Ry(n) = 42, Him, .7» Ry(pu} = 0. and Ry(7) = 21. Since these are all differen, the function {is not continuous at this point. A similar result holds for the profit function. RP) Figure 4.16 Revenue and profit iu ceompetition tons of firm 1 for the model of Bertrand From an economic perspective this discontinuity i also extremely important, since what happens at the point of discontinuity drives the medel to its solution. To see this, consider any particular price, p>, that firm 2 may charge. As Jong as some profit can be male by charging the same price. firm I will never charge a higher priceas that would mean no sales and zero profit. IPfirm 1 charges the same price.136 CHAPTER 4 CONTINUITY OF FUNCTIONS. ary mye) = et Figure 417 eens and poi finsons fn | for aty price etby m2, >4 ‘P| = Pz then the firms share the market, However, if firm 1 charges a slightly lower price than firm 2, twill expture the entire market and its profits will jump ‘higher value. Thus, whatever price firm 2 charges, as long as it exceeds 4, firm will always undercut it in order 1o capture the entire market, It follows that the revenue and profit functions, which are drawn in figure 4.17, are PIM), m= Ps Ried = 4(4)P20- 2p), pi = Pr (43) 0, Pr > Py “The profit function is py(20 —2py) ~ 4202p), mh < Be mip) = 4 (2M20-2p)) —420=291)), ps aa 0. p> If firm 2 sets its price at p> 4, firm | will alway’ have an incentive to se price slightly below firm 2's price in order to capture the entire market, Of cours the same applies to fim 2, Tt will always have an incentive to undereut firm I prive slightly as long as firm 1 is charging a price above $4. Beginning at price above $4 charged by either firm, we will expect the two fiers fo engage in ‘process of undercutting each other's price until the price both fms are chai is S4.and they share the market equally. At this price both firms would eam economic profit and have na incentive to reduce the price further. ‘Thus, by concentrating on the discontinuity of the revenue or profit func Wwe can determine the final outcome of the model, OF course, one must reali that this type of competition will occur only under rather special assumiti Firs, the reselt requires that consumers would immediately switch to be