View original document

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

Economic Review Federal Reserve Bank of Dallas July 1983  1  Will Deregulating Natural Gas Increase Its Price to Consumers?  Stephen P. A. Brown Although it is likely that deregulation will increase the price consumers must pay for natural gas, this increase will occur only if the supply of gas contracts . Decontrol will result in higher wellhead prices and increased production for some categories of gas and lower wellhead prices and decreased production for others. The net effect on the supply of gas and, hence, on the consumer price of gas is not completely clear. However, an analysis of the limited data available suggests that decontrol will probably yield a reduced supply and a higher consumer price.  10  Currency Substitution: The Use of Dollar Coin and Currency in the Texas Border Area of Mexico  William C. Gruben and Patrick j. Lawler  )  Analysis of data on shipments of U.S. currency between banks in the Eleventh Federal Reserve District and banks in Mexico indicates substantial substitution of dollars for pesos in the Texas border area of Mexico during the period from 1973 to 1981. When the risk of a peso devaluation rose, net dollar shipments to Mexico increased considerably. This suggests that border residents weighted more heavily with dollars the currency portfolios they used to conduct their domestic business transactions. By doing so, they limited their capital losses when devaluations occurred. But, collectively, this substitution caused the devaluations to happen earlier by further reducing the demand for pesos.  This publication was digitized and made available by the Federal Reserve Bank of Dallas' Historical Library (  Will Derel~ulating Natural Gas Increase Its Price to Consumersl By Stephen P. A. Brown*  Total deregulation of natural gas prices at the wellhead became a subject of considerable attention in early 1981. Two years later, decontrol has assumed a place of prominence on the legislative agenda. However, consu mer fear that the el im ination of price controls on natural gas will lead to higher prices remains a formidable obstacle to the total deregulation of natural gas prices. Natural gas is the most widely used fuel for home heating. In 1979, 42.5 million households, or 55 percent of all U.S. households, used natural gas for space heating. Another 7 mill ion households, or 9 percent, used natural gas for other purposes. With nearly two-thirds of U.S. households using natural gas, it is no wonder consumers fear that decontrol may result in reduced supply and higher prices. Currently, the wellhead prices of approximately 95 percent of domestic natural gas production are regulated under the Natural Gas Policy Act of 1978 (NG PAl. The 5 percent of gas that is exempt from price controls under the NGPA is very important in determining the effect of complete decontrol of the  * Stephen P. A. Brown is an economist at the Federal Reserve Bank of Dallas. Economic Review/July 1983  natural gas market. Usually, price controls result in a shortage at a price below the market-clearing level. The existence of an unregulated category, one in which the price can rise without limit, allows the natural gas market to clear. Because the NGPA allows the gas market to clear, decontrol will not lead to both a higher consumer price and increased production. Decontrol will yield a higher consumer price only if supply contracts. If decontrol increases supply, a lower price will result. Decontrol sets into motion forces that both expand and contract supply. On January 1, 1985, the NG PA calls for the removal of price controls on production from reserves, known as "new gas," that accounted for about 35 percent of 1982 production. The decontrol measures now before the Congress call for (1) the removal of price controls on production from reserves, known as "old gas," that accounted for about 60 percent of 1982 production and (2) an acceleration of the removal of price controls on new gas. The decontrol of old gas will result in a higher price for such gas, which will call forth additional production from this category. However, with old gas decontrolled, new and unregulated gas will receive prices below those they receive under the NGPA, so production in these two  categories will fall. Although it is impossible to predict confidently whether decontrol will increase or decrease supply, decontrol will likely have a greater impact in reducing the output of new and unregulated gas than in increasing the output of old gas. This prospect means that decontrol can be expected to lead to reduced supply and a higher average price of gas. Furthermore, policies that attempt to attain the benefits of decontrol while imposing new interventionist measures, such as a "windfall profit" tax or a ceiling on the delivered price of gas, will have undesirable consequences. A windfall profit tax would seriously erode the production incentives that decontrol provides for old gas, and a ceiling on the delivered price of gas can be expected to lead to shortages.  Table 1 NATURAL GAS PRODUCTION AND PRICES IN 1982, BY GAS CATEGORY  Scheduled deregulation date  Production (Trillions of cubic feet)  Average wellhead price' per thousand cubic feet  Old gas.  Never  10.184  $1.47  New gas.  January 1,1985  6.402  3.21  (2)  1.016  6.75  17.602  2.41  Category  High-cost gas . All gas  1. Includes applicable severance taxes. 2. Currently deregulated or receives a price greatly in excess of the average for all gas. SOURCE OF PRIMARY DATA: Chase Econometrics.  The natural gas market under the NGPA The NGPA creates over 20 pricing categories of natural gas, to which different ceiling prices are applied. For purposes of analysis here, these categories are lumped into three groups-old gas, new gas, and high-cost gas - as shown in Table 1. As a resu It of Federal regulation, pipelines that transport natural gas from producing areas to consuming areas bring together purchases from all production categories and sell the mixture at a price that represents the average cost of acquiring the gas plus a markup to cover transportation costs (including regulated profit). The price of delivered gas is  (1)  Pd  =  j5  +  IT  = [(.  i.  1=1  P;  Q;) /;=1 /i. Q;]  +  IT.  P is the average wellhead price of gas, IT is the pipeline transportation cost per unit of gas, n is the number of categories in which gas is produced, P; is the wellhead price of gas in category i, and Q; is the quantity of gas produced in category i. In 1982, Pd averaged $4.21 per thousand cubic feet (Md), and the average transportation cost of gas was $1.80 per Md. The 1982 values of P, Pi, and Q; can be found in Table 1. Because the prices for some categories of gas are left unregulated under the NGPA, a market-clearing quantity and average wellhead price are established in the natural gas market. Pipeline companies, acting to satisfy customer demand, bid up the price of gas in unregulated categories (stimulating its pro2  duction) until the average wellhead price and total quantity clear the natural gas market. The price for each regulated category of gas is only bid up until its ceiling price under the NGPA is reached. In the absence of binding price ceilings, competition would assure that each category of gas receives the same price: P1 = P2 = ... = Pn = P. If, however, there is at least one binding price ceil ing, the price of unregulated gas will be above P when a market-clearing equilibrium is achieved. In 1982, as shown in Table 1, unregulated high-cost gas had a wellhead price of $6.75 per Md, while the average wellhead price of all gas was only $2.41 per Md. Those categories with binding ceilings, old gas and new gas, received lower prices per Mcf- $1.47 and $3.21, respectively. The availability of gas to pipelines at prices regulated below those that clear the market, combined with average-cost pricing of delivered gas, leads to a situation in which "cheap" regulated gas in effect subsidizes purchases of higher-priced, unregulated gas. The averaging of cheap pricecontrolled gas with more expensive, unregulated gas holds the price of gas faced by consumers below the marginal cost of gas from the most expensive category. Because it "cushions" the consumer against experiencing the full marginal cost of more expensive gas, the cheaper regulated gas is someFederal Reserve Bank of Dallas  times called cushion gas. Response of the natural gas market to decontrol Decontrol unleashes opposing market forces that both expand and contract supply. Decontrol leads to higher prices for and increased output of old gas and to lower prices for and decreased output of gas in other categories. As shown in Figure 1, an increase in total gas supply will yield a lower wellhead price, and a decrease in supply will result in a higher price. Because natural gas transportation is currently characterized by declining unit costs, the market-clearing price of delivered gas will change more than will the average wellhead price. 1 For analytical purposes, it is convenient to describe supply from each production category as a function of that category's price and supply elasticity. (2)  Q i = Ai P{li  for  i = 1,2, ... , n.  Qi' Pi, and Yli are, respectively, the quantity of category i gas produced, the price of that gas, and the (constant) price elasticity of supply of the gas. Ai is a constant. Note that Yli is greater than zero. The shift in supply resulting from decontrol can be examined through the use of a conceptual experiment in which the production that occurs under the NGPA is compared with that which would occur if each category of gas were to receive the average wellhead price that prevails under the NGPA, P, in place of the price it now receives, Pi' As shown in Figure 1, decontrol shifts the supply curve outward if and only if combined production is increased when each category receives P in place of Pi' Supply shifts inward if and only if combined production is reduced when each category receives P in place of Pi. In performing this experiment to determine the effect of decontrol on supply, it is implicitly assumed that existing supply contracts will not interfere in the adjustment of gas prices to decontrol. Comparison of the experiment with actual production is expressed as (3)  Experiment n  l: A i P'1i  i=1  Actual  > l:n <:  i=1  A i Pi'1i,  1. See the Appendix for further explanation.  Economic Review/July 1983  Figure 1  Supply Response to Decontrol of Natural Gas Prices PRICE (DOLLARS PER THOUSAND CUBIC FEET) SUPPLY DECREASE WITH DECONTROL SUPPLY UNDER NGPA SUPPLY INCREASE WITH DECONTROL DEMAND  o  Q2Q2 Qo Q1Ql  QUANTITY (THOUSANDS OF CUBIC FEET PER YEAR) If decontrol shifts supply outward, the market-clearing wellhead price falls (from P to P1) and the quantity increases (from Qo to Q1). If decontrol shifts ~upply inward, the marketclearing wellhead price rises (from P to P2) and the quantity decreases (from Qo to Q2). The shift in supply can be measured at P. For supply to shift outward, the quantity supplied at P must increase (from Qo to Qi). For supply to shift inward, the quantity supplied at P must decrease (from Qo to Q'). With a downward-sloping demand curve, the shift in supply is greater than the change in the market-clearing quantity.  where P and Pi refer to values that prevail under the NGPA. If an equality holds in expression 3, decontrol leaves supply unaltered (over the relevant range) because total production would be unaltered at P. Consequently, the market-clearing wellhead price of gas will remain unchanged. However, if the lefthand side of (3) is greater than the right-hand side, decontrol would increase supply and lead to a lower market-clearing price of gas than would occur under the NGPA. On the other hand, if the left-hand side of (3) is less than the right-hand side, decontrol would decrease supply and lead to a higher marketclearing price of gas. Decontrol increases (decreases) supply if the experiment yields an increase in production for 3  Table 2 SUPPLY RESPONSES TO CHANGES IN PRICES FROM Pi TO P, BY GAS CATEGORY Estimates for 1982  Category  Supply elasticities  With low elasticity  With midpoint elasticity  With high elasticity  Trillions of cubic feet  Old gas ...... .  .05 to ,20  0,240  0,649  ',058  New gas .. .  0.3 to 0,8  -,528  -,934  -1.312  High-cost gas ...  1,2 to ',5  -.721  -.763  -.799  NOTE:  ~i  is the price that prevails under the NGPA for each gas category. P is the average price of gas that prevails under the NGPA. SOURCES OF PRIMARY DATA: Chase Econometrics. Data Resources, Inc. U.S. Department of Energy.  categories with ceiling prices below P that is greater (less) than the decrease in production experienced in categories with prices above P while price controls are in effect. This can be seen more clearly by subtracting Ai P i'1i from both sides of expression 3, substituting Qi/ P i'1i for A i' and reducing terms to obtain  (4) where Qi' P, and Pi refer to values prevailing under the NGPA.  Estimating the response to decontrol A precise estimate of the effect that decontrol has on gas supply cannot be constructed from available data, However, a range of supply responses can be estimated with expression 4, The actual change in supply occurring in response to decontrol can be expected to fall within the estimated range, With the static formulation of the natural gas market contained in expression 4, a range of longrun price elasticities of reserves can be used to predict the effect of decontrol on production in the first year for which long-run adjustment is complete, For analytical purposes, long-run adjustment is assumed to take place in the same year in which 4  decontrol commences, Estimates of the supply elasticities were obtained as follows, The U,S, Department of Energy (DOE) has estimated that reserves of old gas would be increased by 2.5 to 11 trillion cubic feet (Tef) if total decontrol were to take place,2 I n view of the 1981 reserve-to-production ratio of 10.4,3 the DOE estimate implies that total decontrol would increase the production of old gas by 0,240 to 1,058 Tef in the first year of decontrol (assuming long-run adjustment of "old" reserves to decontrolled prices), Given that the production increase of 0,240 to 1,058 T ef represents the supply response to raising the price of old gas from $1.47 per Mef to $2.41, the DOE estimate would have translated to an elasticity of supply of old gas between 0,05 and 0,20 in 1982, While elasticities of supply for new and high-cost gas are not known precisely, values of 0.3 to 0,8 for new gas and 1,2 to 1.5 for high-cost gas are considered within reason by economists working in the area of natural gas supply,  2. As reported in Oil Daily, 7 April 1983, 1. 3. At the time of this writing, 1981 data on reserves were the latest available. See DeGolyer and MacNaughton, Twentieth Century Petroleum Statistics, 1982 (Dalias, Tex,: DeGolyer and MacNaughton, 1982), 75.  Federal Reserve Bank of Dallas  Table 3  ESTIMATED SHIFTS IN GAS SUPPLY AT  P  First year of decontrol  1982  1983  1984  1985  1990  Trillions of cubic feet  Scenario1 ..  -1.871  -1.853  -1.626  -1.406  -1.300  Scenario 2 .  -1.048  -1.019  -.963  -.667  -.516  Scenario 3.  -.191  -.154  -.103  .086  .277  Assumptions  Scenario1  Old gas has its low-elasticity value, while new gas and high-cost gas have their high-elasticity values.  Scenario 2 .  Old gas, new gas, and high-cost gas have their midpoint-elasticity values.  Scenario 3 . ..  Old gas has its high-elasticity value, while new gas and high-cost gas have their low-elasticity values.  NOTE:  P is  the average price of gas that prevails under the NGPA. A negative figure denotes an inward supply shift.  These elasticities are used to estimate changes in supplies of old, new, and high-cost gas that would have occurred in 1982 had each of the categories received $2.41 per Md - and not $1.47, $3.21, and $6.75, respectively. These estimates, shown in Table 2, can be used to identify the range within which the change in supply brought about by decontrol could have been expected to fall. Even with the maximum expected response of old gas to decontrol and minimum expected responses of new and highcost gas, decontrol in 1982 would have shifted 1982 supply inward 0.191 Td at $2.41 per Md and pushed the market-clearing price upward. For the midpoint-elasticity values, the inward shift is estimated at 1.048 T d. As shown in Table 3, the maximum inward shift that could have been expected had decontrol occurred in 1982 is estimated at 1.871 T d. As one looks from 1983 toward 1990, the picture begins to change. As time depletes the current reserves of old gas and this category takes a smaller share of the market,4 the average wellhead price of gas moves closer to prices of the more expensive categories. This decline in market share diminishes the effect of decontrol on the production of new and high-cost gas. At the same time, the expected Economic ReviewlJuly 1983  additions to old-gas reserves resulting from decontrol remain unchanged. The upshot is that as time advances, decontrol becomes more attractive (or less unattractive) from the standpoint of the change in consumer prices. An outward supply shift enters the range of likely responses by 1985, and the maximum expected inward shift is reduced. Nevertheless, the midpoint case (Scenario 2) continues to yield an inward shift through 1990, taking a positive value at some date beyond 1990. The decontrol of old gas prior to 1990 is likely to result in higher gas prices, but there is a possibility that decontrol before that year will lead to lower gas prices. The possibility that decontrol will lead to lower gas prices increases over time.  Shifting controls from the wellhead to deliveries of gas It might seem that a policy that shifts price controls from the wellhead to the delivery market could pro-  4. With a small contribution from additional decontrol scheduled for 1987 under the NGPA, depletion is expected to reduce the market share of gas remaining under price controls to 20 percent by 1990.  5  Figure 2 II Decontrol" of Natural Gas with an Effective Ceiling on the Average Price  PRICE (DOLLARS PER THOUSAND CUBIC FEET)  will get less gas than under the NGPA, and there will be a shortage. Similar shortages do not arise under the NGPA because the price of gas in uncontrolled categories is free to rise until a marketclearing quantity and average price are reached. A shortage of natural gas induced by a price ceiling could lead to higher overall energy costs because more expensive substitutes must be utilized. Temporary suspensions of delivery (service interruptions) being the most visible evidence of natural gas shortages, it is commonly assumed that industry bears the costs of a shortage. However, previous curtailments have been accompanied by the less visible denial of hookups- both residential and industrial- indicating that at least some residential consumers could face higher energy costs under price ceilings.  Windfall profit taxes o  Q2  Q1  Qo  QUANTITY (THOUSANDS OF CUBIC FEET PER YEAR) If decontrol shifts supply inward (from So to 5,), the marketclearing price of natural gas rises (from to P,) while the market-clearing quantity falls (from Qo to Q,). If a ceiling price on delivered gas is then established at P to keep consumers from facing higher prices, a shortage of (Qo-Q,) will develop.  P  tect the consumer against higher gas prices while retaining some of the benefits of decontrol. Nevertheless, placing a price ceiling on delivered gas will either have no effect on the natural gas market or, as is more likely, lead to a shortage. If the lifting of price controls at the wellhead shifts supply outward, reducing the market-clearing average price, a ceiling price established at the NGPA-controlled average price would be greater than the decontrolled average price. Hence, the ceiling price would be of no consequence. On the other hand, if the lifting of wellhead price controls shifts supply inward, increasing the marketclearing average price, a ceiling price established at the NGPA-controlled average price would be less than the decontrolled market-clearing price. As shown in Figure 2, the price of gas would be held down to the NGPA-controlled average price. And while consumers might pay no higher prices, they 6  Increased producer profits resulting from decontrol are viewed as undesirable windfalls in some circles. "Windfall profit" taxes have accompanied U.S. oil price decontrol and have been discussed in conjunction with possible decontrol of old gas. Unfortunately, stripping "windfall profits" from producers without reducing their production incentives appears impossible. If a windfall profit tax on natural gas is similar in structure to that on crude oil, producers of old gas would pay an excise tax on the difference between its free-market price and the controlled price that would have prevailed under the NGPA. To capture the effect of "windfall profit" taxes on supplies, the test for the direction of total supply change, expression 4, is modified:  (5)  n  ~  i=1  Q(  [(p - t. max[O (P -  P.)l)'1i  ' I  Pi  in which t is the windfall profit tax rate. All other variables are previously defined. The effects of the windfall profit tax were examined at a 50-percent tax rate. Far from preventing producers of old gas from benefiting at the expense of consumers, a windfall profit tax actually benefits the Government at the expense of consumers and producers of old gas. It seriously erodes the production incentives for old gas brought about by decontrol. It shifts supply inward, yielding higher conFederal Reserve Bank of Dallas  Table 4  ESTIMATED SHIFTS IN GAS SUPPLY AT P, GIVEN A 50-PERCENT "WINDFALL PROFIT" TAX RATE First year of decontrol  1982  1983  1984  1985  1990  Trillions of cubic feet  Scenario 1 .  -1.969  -1.959  -1.733  -1.513  -1.404  Scenario 2.  -1.338  -1.309  -1.256  -.960  -.803  Scenario 3.  -.668  -.632  -.586  -.398  -.202  Assumptions  Scenario 1 ..  Old gas has its low-elasticity value, while new gas and high-cost gas have their high-elasticity values.  Scenario2.  Old gas, new gas, and high-cost gas have their midpoint-elasticity values.  Scenario 3.  Old gas has its high-elasticity value, while new gas and high-cost gas have their low-elasticity values.  NOTE:  P is the average price of gas that prevails under the NGPA. A negative figure denotes an inward supply shift.  sumer gas prices than would prevail without the tax. As shown in Table 4, the tax increases the likelihood that decontrol will lead to higher prices. At the same time, producers of old gas will obtain a price lower than would prevail without the tax-though higher than prices prevailing under the NGPA. And with a windfall profit tax rate of 50 percent on decontrolled old gas, the Federal Government could have obtained revenues of approximately $5 billion in 1982. Concluding remarks  The response of total gas supply to price decontrol is the sum of production responses in individual NGPA categories. If the additional gas brought forward from newly decontrolled old-gas categories is greater than the production decrease in higher-cost categories that face lower prices with decontrol than under the NGPA, decontrol shifts total supply outward, and the consumer price of gas is lower than under the NGPA. On the other hand, if the additional gas brought forward from newly decontrolled old-gas categories is less than the production decrease in higher-cost categories that face lower prices with decontrol than under the NGPA, deconEconomic Review/July 1983  trol shifts total supply inward, and the consumer price of gas is higher than under the NGPA. Given probable ranges of the supply elasticities of old, new, and high-cost gas, it is impossible to predict whether decontrol will increase or decrease consumer gas prices. However, decontrol before 1990 is likely to lead to a reduced gas supply and higher consumer gas prices, but as time advances, decontrol becomes less unattractive from the point of view of consumer prices. At some time beyond 1990, decontrol would likely yield a lower gas price than if the NGPA is maintained. Combining "decontrol" with new interventionist measures to protect consumers from higher prices or to prevent "windfalls" from accruing to producers of old gas will have undesirable results. The efficiency gains of natural gas decontrol can only be obtained if consumers face the risk of higher prices and old-gas producers are afforded freemarket production incentives. When evaluating complete decontrol, it is important to keep in mind that the NGPA calls for the removal of price controls on new gas in 1985. The same analytical procedures used to examine the effects of total decontrol can be used to analyze the 7  consequences of partial decontrol under the NGPA. Given the probable ranges of supply elasticities of new and high-cost gas, it would appear that natural gas prices faced by consumers will be lower in 1985 than if the NGPA did not provide a decontrol of new gas in that year. Advancing the date by which partial decontrol is brought about appears beneficial to gas consumers. Both of these results are dependent on the assumption that existing supply contracts will not interfere with the adjustment of gas prices to new market conditions. For economic analysts, existing contracts between pipelines and producers are perhaps the most troublesome aspect of the natural gas market. Provisions in many new-gas supply contracts appear to call for sharply increased prices following the scheduled decontrol in 1985. Under these contracts the price of gas from "new" reserves generally moves upward to a price equivalent, in British thermal units, to that for distillate fuel oil or to the  average of the three highest prices paid for gas in the same producing locale. The effect of these contracts on prices in a differing market environment is difficult to foresee. The author has taken the approach that existing contracts between producers and pipelines are not a significant obstacle to the adjustment of gas prices to changed conditions in the market. However, some analysts have concluded that pricing provisions in contracts negotiated in the late 1970's will lead to sharply higher gas prices when the NGPA decontrols new gas in 1985. If these analysts are correct, decontrol legislation that also voids pricing provisions in currently existing contracts is likely to lead to lower consumer gas prices than would prevail under the NGPA, but the voiding of these pricing provisions would be the significant factor in such legislation and not the decontrol of old-gas prices.  Appendix Some Mathematics of Natural Gas Demand In response to decontrol, the wellhead price of gas will move in the same direction as the delivered price, though its change will be smaller than that exhibited by the delivered price. The wellhead demand price of gas, P, is the difference between the delivered demand price of gas and the unit transportation cost: (A.1)  P  = Pd(Q, Px'  Y) -  n(Q, C).  The delivered demand price, Pd , is a function of the quantity of gas that must be absorbed, Q; the prices of other goods, Px ; and a national income measure, Y. The unit transportation cost, n, is a function of the quantity of gas transported, Q, and the cost of transportation inputs, C. The change in the wellhead demand price resulting from an increasing (or decreasing) quantity reflects both consumers' willingness to absorb additional 8  (reduced) supplies and corresponding changes in unit transportation costs: (A.2)  aplaQ  =  aPdlaQ - anlaQ.  Natural gas being a commodity that obeys the law of demand, Pd l Q is negative. However, because natural gas pipelines have very high fixed costs, excess capacity into the foreseeable future, and a regulated rate of return, natural gas transportation appears to have decreasing unit costs, yielding a negative anI aQ. Given negative values of ~PdlaQ and anlaQ, it can be concluded that aPlaQ is greater than aPdlaQ.lf aPlaQ is negative, the wellhead price will have a smaller response to decontrol than will the delivered price. There remains the possibility that aPlaQ is positive. In this case the wellhead price of gas will move in the opposite direction from the delivered  a  a  Federal Reserve Bank of Dallas  price in response to decontrol. Delivery market demand is represented as (A.3) in which B is a constant that captures the influence of all variables other than Pd on demand and fJ is the price elasticity of demand. The average unit tr~nspor­ tation cost is the sum of average fixed costs and average variable costs: (A.4)  TI  = TI f  +  TIv'  Average fixed transportation costs, TI f , equal total fixed costs divided by quantity, TFC/Q, and average variable costs are assumed constant, (0 TIv/ 0 Q = 0). Using oPd/oQ and OTI/oQ from (A.3) and (A.4), substituting (P/ld/Q)-1 for Band (TIf Q) for TFC and then reducing terms, (A.2) can be rewritten as (A.S)  + TIf Q -1 Pd + TIf).  o P /0 Q = fJd -1 PdQ -1  =  Economic Review/July 1983  Q-1'(fJd -1  Because Q is positive, the sign of 0 P/0 Q is determined by_ the sign of (fJd -1 Pd + TI f ). Rearrangement yields (0 P / 0 Q < 0) if and only if (fJd > - P /TIf). Even if the 1982 transportation cost of $1.80 per Mcf represented only f~xed costs, fJd would have to be below -2.34 for oP/oQ to be positive. American Gas Association data reveal that TIf was less than $1.80 per Mcf in 1982.' Furthermore, from Douglas Bohi's survey of econometric studies of energy demand,2 it is safe to infer that the composite U.S. elasticity of demand for gas is greater than - 2.34. The upshot is that though oP/oQ is greater than oPd/oQ, it is still negative.  1. Gas Facts: 1981 Data (Arlington, Va.: American Cas Association, 1982), 143-86. 2 Douglas R. Sohi, Analyzing Demand Behavior: A Study of Energy Elasticities (Baltimore: Johns Hopkins University Press for Resources for the Future, 1981),  92-113.  9  Currency Substitution: The Use of Dollar Coin and Currency • In the Texas Border Area of Mexico By William C. Gruben and Patrick j. Lawler*  Events in the early 1980's have dramatically illustrated the importance of economic conditions in Mexico to prosperity in Texas border cities. In 1980 and 1981, Mexico's economy was expanding rapidly, and sales to Mexicans by merchants on the U.S. side of the border increased sharply. In 1982 the peso was devalued, and Mexico's economy slowed. These developments were accompanied by an abrupt decline in retailing and a sharp rise in unemployment along the U.S. side of the border. Underlying these widely publicized developments were many relatively unnoticed activities. During the 1970's and until the initiation of exchange restrictions in the fall of 1982, armored cars carrying U.S. and Mexican currency traveled daily between Texas and Mexican banks near the Rio Grande. The armored cars did not simply return U.S. tourist dollars and pesos of Mexican shoppers to their home countries; they often took U.S. currency to Mexico. Some Texas border banks shipped as much as $300,000 to $500,000 per day to Mexican banks for periods of time during 1980 and 1981.'  * William C. Gruben and Patrick J. Lawler are senior economists at the Federal Reserve Bank of Dallas. 10  These shipments reflect an important aspect of life along the Texas-Mexico border: residents have a choice of two currencies, and they are alert to symptoms of disequilibrium in the market for these currencies. It is conceivable, of course, that the shipments of dollars to Mexico are entirely attributable to the needs of Mexican residents for purchasing goods in the United States. But Mexicans might also be using dollars for purely internal trans-  1. Each quarter, the Federal Reserve Bank of Dallas surveys 11 Texas banks on the Mexican border with regard to their shipments to and receipts from banks in northern Mexico of both dollar and peso currency. The results of this survey, which is performed using Form FR 2465, "Receipts and Shipments of United States and Mexican Currency Between United States Banks and Mexico," are sent to the U.S. Department of Commerce. The Federal Reserve Bank of San Francisco (Twelfth Federal Reserve District) also conducts such a survey, collecting reports from 10 banks in Arizona and California. According to the July 22, 1982, Federal Reserve memorandum entitled "Proposal for Extending Authorization for Form FR 2465, Receipts and Shipments of United States and Mexican Currency Between United States and Mexico," the survey samples are estimated to capture about 80 percent of the currency shipments in the Eleventh District and about 95 percent in the Twelfth District.  Federal Reserve Bank of Dallas  actions as substitutes for potentially depreciating pesos or, possibly, replacing peso hoards with dollar hoards. Residents along the border are generally believed to be well informed about developments that might affect the peso/dollar exchange rate. Because holding dollar currency as an alternative exchange medium is a cheap and convenient way for Mexican residents to protect a portion of their wealth, some part of the large amount of dollars shipped to Mexican banks is probably attributable to concern about deval uation. To the extent that this is the case, the availability of dollars to Mexicans near the border can be viewed as a factor that limits the options open to Mexico's monetary authorities. This article analyzes a previously neglected data source: shipments of currency between Texas and Mexican banks. Both visual examination of time series data on currency shipments and a regression equation based on a model of dual currency demand show that during the period from 1973 to 1981, net shipments of dollars to Mexico were systematically related to a measure of concern about the prospect of a peso devaluation. The evidence supports earlier studies by others that found dollar deposits were substituted for peso deposits when the probabilities of peso devaluation increased. This willingness of the Mexican public to use dollar currency and deposits may limit the ability of the Mexican government to follow an independent monetary policy successfully and increase the discipline of fixed or closely managed exchange rates.  Currency substitution and the U.S.-Mexico border Use of multiple moneys has occurred in Mexico throughout the 20th century. During the administration of General Porfirio Dfaz, circulation of foreign currency was sufficiently common to lead the government to prohibit it in 1905. In the revolution following the overthrow of Dfaz in 1910, at least 21 varieties of paper money circulated, along with a large supply of counterfeit currency. By 1917, most Mexicans accepted nothing but gold and silver coins. U.S. gold coins circulated in Mexico from that time until the early 1930'S.2 More recently, checkable deposits denominated in U.S. dollars became an important component of Mexico's money stock. This phenomenon has been referred to as "dollarization."3 Over the 1973-81 Economic ReviewlJuly 1983  period, the proportion of the peso value of checking deposits in Mexico that were denominated in dollars ranged from 5 to 12 percent of total checking deposits. Dollar-denominated checking deposits have fallen below 3 percent of total Mexican checking deposits in the wake of the country's effective prohibition of such deposits in 1982. Even this 3 percent cannot serve as a medium of exchange because, although the accounts are still carried as dollar-denominated, checks may only be drawn in pesos. On May 13,1983, dollar-denominated savings accounts were reinstated but only for certain types of businesses. In recent years, economists have been addressing phenomena such as these dollar-denominated checking accounts under the heading of "currency substitution."4 Although the term is used in a variety  2. Two good articles dealing with Mexican monetary history are Manuel Cavazos Lerma, "Cincuenta anos de politica monetaria," in Cincuenta anos de banca central. compo Ernesto Fernandez Hurtado (Mexico City: Fondo de Cultura Economica and Banco de Mexico, 1976), 55-122, and Raul Martinez Ostos, "EI Banco de Mexico, 1925-1946," app. 1 to M. H. de Kock, Banca central, 3d ed. (Mexico City: Fondo de Cultura Economica, 1955), 367-424. 3. For examples of literature on the dollarization problem in Mexico, see Leroy O. Laney, "Currency Substitution: The Mexican Case," Voice of the Federal Reserve Bank of Dallas, January 1981, 1-10; Guillermo Ortiz, "Currency Substitution in Mexico: The Dollarization Problem," Journal of Money, Credit, and Banking 15 (May 1983): 174-85; Guillermo Ortiz Martinez, "La dolarizacion en Mexico: Causas y consecuencias," Monetaria 5 (October-December 1982): 439-64; Guillermo Ortiz and Leopoldo Solis, "Currency Substitution and Monetary Independence: The Case of Mexico," in The International Monetary System Under Flexible Exchange Rates: Global, Regional, and National. ed. Richard N. Cooper, Peter B. Kenen, Jorge Braga de Macedo, and Jacques van Ypersele (Cambridge, Mass.: Ballinger Publishing Company, 1982),217-33; and Guillermo Ortiz and Leopoldo Solis, "Inflation and Growth: Exchange Rate Alternatives for Mexico," in Exchange Rate Rules: The Theory, Performance and Prospects of the Crawling Peg, ed. John Williamson (New York: St. Martin's Press, 1981), 327-46. 4. Arguments concerning problems of currency substitution appear in the following articles by Marc A. Miles: "Currency Substitution, Flexible Exchange Rates, and Monetary I ndependence," American Economic Review 68 (June 1978): 428-36; "Currency Substitution: Perspective, Implications, and Empirical Evidence," in The Monetary Approach to International Adjustment, ed. Bluford H. Putnam and D. Sykes Wilford (New York: Praeger Publishers, Praeger Special Studies,  11  of contexts, it rarely refers literally to substitution of foreign paper and coin currencies for the domestic issues. Currency substitution usually means the holding of non-interest-bearing deposits denominated in a foreign currency. Use of accounts denominated in a foreign currency could arise merely for convenience, but economists' interest is normally focused on currency diversification as a response to uncertainty about the future course of foreign exchange rates. While such uncertainty is usually associated with flexible exchange rates, socalled fixed exchange rates are subject to pressure and are adjusted from time to time. I n the case of Mexico, periods of essentially fixed rates have been interrupted occasionally by devaluations. Thus, on occasions when Mexican citizens believe that their government may be losing the ability to manage its exchange rate, people are likely to hold more dollars in order to limit their losses if a devaluation should occur. Because this response will, to some degree, increase pressure on the exchange rate, dollarization is one of the forces restraining the government's ability to cause Mexico's inflation rate to depart radically from that of the United States without devaluing the peso. Dollarization is particularly evident in northern Mexico. Frequent transactions between parties on opposite sides of the U.S.-Mexico border allow Mexican residents to acquire and dispose of dollars conveniently. The Mexican border cities are dotted with maquiladoras, assembly plants specializing in production for export using materials chiefly from the  1978), 170-83; and "Currency Substitution: Some Further Results and Conclusions," Southern Economic Journal 48 (J uly 1981): 78-86. There is evidence that currency substitution is not as significant as Marc Miles perceives, however. For additional discussions of currency substitution in a context of flexible exchange rates, see Arturo Brillembourg and Susan M. Schadler, "A Model of Currency Substitution in Exchange Rate Determination, 1973-78," International Monetary Fund Staff Papers 26 (September 1979): 513-42; Lance Girton and Don Roper, "Theory and Implications of Currency Substitution," Journal of Money, Credit, and Banking 13 (February 1981): 12-30; Michael D. Bordo and Ehsan U. Choudhri, "Currency Substitution and the Demand for Money: Some Evidence for Canada," Journal of Money, Credit, and Banking 14 (February 1982): 48-57; and Leroy O. Laney, Chris D. Radcliffe, and Thomas D. Willett, "International Currency Substitution by Americans Is Not High: A Comment on Miles," Southern Economic Journal. forthcoming.  12  United States. Many of these plants are closely related to factories just across the border that will carry out additional procedures following the assembly operations in Mexico. The ties between parties on the two sides of the border are further strengthened by the large percentage of wholesale and retail sales in some U.S. cities along the Mexican border that are made to Mexican citizens. 5 I ndeed, it has been argued that Mexico's northern border can almost be considered a currency area separate from the rest of Mexico. 6 Dollar checking deposits, although common throughout Mexico until late 1982, were more heavily concentrated along the northern border than in the central and southern parts of the country. At the end of 1981, as shown in the table, dollar-denominated demand deposits accounted for twice as large a share of total demand deposits in the four largest Mexican border cities across from Texas (Ciudad Juarez, Matamoros, Nuevo Laredo, and Reynosa) as in the country as a whole. The 12 cities in Mexico along the U.S. border for which banking data are reported accounted for about 6 percent of total demand deposits in Mexico and for nearly 18 percent of dollar-denominated demand deposits. The six Mexican states bordering on the United States were responsible for 20.9 percent of total demand deposits in Mexico and for 30.5 percent of dollar-denominated demand deposits. For Mexico as a whole, the available evidence indicates that exchange rate risk is an important contributor to variation in dollar-denominated deposits. Surges in dollarization preceded major devaluations  5. For further discussions of these phenomena, see, for example, Edmundo Victoria Mascorro, "Caracteristicas del desarrollo economico de la franja fronteriza norte de Mexico," Natural Resources Journal 22 (October 1982): 823-45; Niles Hansen, The Border Economy: Regional Development in the Southwest (Austin: University of Texas Press, 1981); Oscar J. Martinez, Border Boom Town: Ciudad Juarez Since 1848 (Austin: University of Texas Press, 1978); and Victor Urquidi and Sofia Mendez Villarreal, "Economic Importance of Mexico's Northern Border Region," in Views Across the Border: The United States and Mexico, ed. Stanley R. Ross (Albuquerque: University of New Mexico Press in cooperation with the Weatherhead Foundation, 1978), 141-62. 6. This argument is offered by Ortiz, La dolarizacion en Mexico, 451. In support of his contention, Ortiz goes on to say that prices are quoted in dollars, the majority of payments are made in dollars, and the peso, although accepted on both sides of the border, is less demanded than the dollar.  Federal Reserve Bank of Dallas  SELECTED MEXICAN BANK DEPOSITS, DECEMBER 31,1981 (Values in millions of pesos)  Area  Republic of Mexico .. Largest cities bordering on the Eleventh District Ciudad Acuna, Coahuila. Ciudad Juarez, Chihuahua .. Matamoros, Tamaulipas . Nuevo Laredo, Tamaulipas. Piedras Negras, Coahuila Reynosa, Tamaulipas Mexican states bordering on the Eleventh District Chihuahua Coahuila. Nuevo Leon ......... Tamaulipas Other Mexican states bordering on the United States Baja California Norte Sonora ..........  Total value of all deposits  Dollar-denominated deposits As percent of all deposits Value  Total value  Checking deposits Dollar-denominated As percent of all checking deposits Value  1,505,222.1  360,795.3  24.0  354,668.2  42,892.6  12.1  1,023.1 14,500.5 9,108.8 8,891.3 3,279.6 8,271.5  382.1 6,281.5 2,581.8 3,114.8 1,065.0 2,643.2  37.3 43.3 28.3 35.0 32.4 31.9  144.8 2,772.5 1,422.7 1,473.7 523.0 1,499.7  54.3 739.7 376.3 403.9 163.1 374.8  37.5 26.7 26.4 27.4 31.2 25.0  39,332.9 38,556.1 100,306.2 52,649.8  10,111.9 6,703.1 17,073.0 12,251.3  25.7 17.4 17.0 23.3  9,846.7 9,726.3 22,974.1 10,870.9  1,124.4 1,138.3 2,553.1 1,630.4  11.4 11.7 11.1 15.0  55,331.3 40,357.0  29,186.6 10,390.5  52.7 25.7  10,155.7 10,386.9  4,977.5 1,671.8  49.0 16.1  SOURCE OF PRIMARY DATA: Banco de Mexico.  Economic Review/July 1983  13  Chart 1  Interbank Transfers of Dollars Between Eleventh District and Mexico MILLIONS OF DOLLARS 300 ~------------------------------------------------------------, (QUARTERLY) GROSS SHIPMENTS TO MEXICAN BANKS  200 4  ----.------ -----, ,.,...I '  GROSS RECEIPTS FROM MEXICAN BANKS,' \  100  -----~,  ~  ,  ~ .....\ ,~ ,~\"oJ  , ,  -100 -200  ~-L  ____-L____-L____-L____-L____-L____-L____-L____-L____  -L~  1977  in 1954, 1976, and 1982.7 Because this observation is based primarily on visual inspection of the time series of dollar-denominated demand deposits rather than on any formal, rigorous analysis, one cannot infer cause and effect with absolute confidence. Mexicans, especially those near the border, may accumulate dollar balances in anticipation of transactions with U.S. businesses or individuals, so concern about devaluation is not the only factor contributing to increases in dollarization. Because a divergence in inflation rates and, therefore, price levels is an important contributor to pressure on the exchange rate, purchases from the United States could be expected to rise at the same time as concern about devaluation. The currency shipments data provide a unique opportunity to examine currency substitution in the literal sense of the term. Information on quantities of foreign currency circulating is not available for any country because currency holdings are so widely dispersed. The interbank shipments of dollars and pesos do not allow estimation of the quantity of currency circulating in Mexico at anyone time;  7. See Laney, "Currency Substitution: The Mexican Case." 14  however, this series does provide indirect information on changes in currency in circulation, and these changes are sufficient for an analysis of the determinants of the quantity of dollar currency demanded. If concern about devaluation is an important determinant of the demand for dollars in the Texas border area of Mexico, then net interbank shipments should vary predictably with changes in pressure on the exchange rate. Convincing evidence that holdings of dollar currency in Mexico respond to exchange rate pressure is not easily obtained. Visual inspection of the interbank shipments series in Chart 1 shows that net shipments of dollars to Mexico do indeed rise in months preceding devaluations. Although this relationship is consistent with what one would expect if Mexicans responded to exchange rate risk, other factors that also vary with such risk may account for the relationship. Visual inspection also shows that purchases of U.S. goods by border Mexicans are highly correlated with dollar shipments (Chart 2). Thus, the international transactions motive for holding dollars may explain most of the variation in the net shipments series. The next two sections of this article examine these issues in more detail. First, the behavior of Federal Reserve Bank of Dallas  Chart 2  International Border Expenditures by Mexicans and Dollar Shipments from Eleventh District Banks to Mexican Banks MILLIONS OF DOLLARS  500 r-------------------------------------------------~~~~~ (QUARTERLY)  ,"  400  ,.  ,,;'  ;'  300  BORDER EXPENDITURES "  ,-,'" " "  200 100  .,.,.,_,_,'-1_1_'._  ~~  ",,--,~  series on border transactions and Mexican national income will be discussed in relation to recent Mexican economic history and the behavior of the interbank shipments data. Then, a formal statistical analysis of the issue will be presented. The Mexican economy, 1973-1981 During the 1946-72 period, Mexico enjoyed relative economic stability, characterized by annual growth rates of 3.8 percent for consumer prices, 11.5 percent for the narrowly defined money stock, and 6.6 percent for real gross domestic product (GOP). The period of stability ended in 1973, as the consumer price index rose 12.1 percent and the money stock jumped 22.4 percent. An increase in world inflation and the decision of the Mexican government to resist world recessionary pressures through rapid growth of public expenditures helped bring on the shift in the country's economic performance. For the 1973-76 period, Mexico's rate of inflation was double that of the United States. After significant oil strikes beginning in the early 1970's, Mexico Economic Review/July 1983  ",  #  ;'  "  ,,'  DOLLAR SHIPMENTS  ,  ~'_._I_I#  had decided to use the income it expected to receive from oil exports to fund a broad-based set of investments. But as Mexico increased public expenditures to pursue these goals, it began to rely heavily on external sources of credit and to print money at an increasingly rapid rate. During this period, a continuing differential between the U.S. and Mexican inflation rates, coupled with a fixed exchange rate, induced Mexican buyers to purchase a rising proportion of goods and services abroad and exacerbated pressure on the peso/dollar exchange rate. The effects of all these events were felt strongly on both sides of the border. During 1973-75, border income to businesses and individuals from the sale of goods and services to purchasers abroad rose at an average annual rate of 13.7 percent. However, over the same period, Mexican border expenditures for goods and services abroad rose at a much higher rate of 19.3 percent. Income from border transactions net of expenditures has historically proved a source of foreign exchange in Mexico. Mexican in15  ternational expenditures in border transactions tend to be less than such income, and this tendency continued during the period. However, between December 1973 and December 1975, Mexican border expenditures for goods and services abroad as a proportion of Mexican border income from international sales rose from 61 percent to 69 percent. The higher rate of increase for expenditures in international transactions by border households and institutions than for income from such transactions was accompanied by rising demand for interbank shipments of U.S. currency to northern Mexico. Currency shipments from banks in the Eleventh Federal Reserve District to Mexican banks rose steadily, while receipts of dollars by Eleventh District banks from Mexican border banks declined. The portion of the rising demand for currency due to increased purchases of U.S.-produced goods and services is, however, not reflective of currency substitution. Such currency demand reflects only the attractiveness of acquiring U.S. currency from Mexican banks in order to purchase U.S. products. Furthermore, the expectation of a devaluation in the near future is likely, in and of itself, to induce Mexicans to hurry to buy U.S. products now rather than later. The same expectation can also motivate currency substitution. Pressure on the peso increased until August 31, 1976, when the Mexican monetary authorities floated the peso. The exchange rate rose from 12.5 pesos to the dollar in August to 20-22 pesos per dollar in September and to 24-25 by November. Inflation subsided for a while after Mexico adopted a stabilization program following the devaluation. However, with time, resumption of growth-oriented policies was reflected in large increases in prices and in the money stock. For the 1979-81 period the annual average increase in the Mexican consumer price index exceeded 24 percent, more than double that of the United States. More increases in foreign debt occurred. Beginning in the second quarter of 1977, the Mexican government had allowed the peso/dollar exchange rate to fluctuate within very narrow margins. In 1980 the Mexican government began to allow the peso to slide against the dollar but not by nearly enough to bring the two currencies in line with their relative purchasing powers. These pressures again were reflected on the border, where during 1979-81 the average annual 16  rate of increase in income from international border  transactions was 22.8 percent, compared with an annual rate of increase of 26.1 percent for expenditures. Mexican expenditures in international border transactions as a proportion of income from such transactions rose from 74 percent in the fourth quarter of 1977 to 84 percent in the fourth quarter of 1979 and to 102 percent in the fourth quarter of 1981. Also, after declines in shipments of dollars to Mexican banks by Eleventh District banks in 1977 and 1978, and with increases in receipts of dollars moving in the opposite direction, District bankers began to see rising demand for dollars by banks in northern Mexico in 1979. This pattern continued through 1980 and 1981. By 1981, it was clear that Mexico was facing very serious econom ic problems. Mexico's 1981 cu rrent account deficit was $13.9 billion, compared with $7.7 billion in 1980 and $5.5 billion in 1979. The public sector fiscal deficit, which had constituted 7.3 percent of Mexican gross domestic product in 1980, rose to 14.8 percent in 1981. Subsequently, the Mexican government again allowed the market mechanism to take its course in the foreign exchange market. At the end of January 1982, the exchange rate was 26.6 pesos per dollar; a month later, it reached 44.6. By the end of the year, the market rate had soared to 148.5.  A model of interbank currency transfers between southern Texas and northern Mexico A model of dual currency demand was used to derive a suitable equation for estimating the effects of prospective peso devaluation on the demand for dollar currency in the four Mexican states that border on the Eleventh District. The model is briefly described here; mathematical detail is left to the Appendix. Assume that border Mexicans demand currency services but are prepared to receive these services from either dollars or pesos. Although pesos serve more efficiently for many sorts of transactions while dollars may serve more efficiently for others, the two currencies are at least somewhat substitutable. If Mexican households and businesses near the border attempt to minimize the cost of obtaining transaction services that currency provides, then as the opportunity cost of holding pesos rises relative to the opportunity cost of holding dollars, they will shift their currency portfolios to include more Federal Reserve Bank of Dallas  dollars and fewer pesos. Empirical evidence of such shifts would support the conclusion that currency substitution does occur. The model as outlined so far provides currency stock demand functions. Because there are no stock data for dollars used in Mexico, the model must be adapted to use the partial information about flows provided by data on cross-border interbank shipments of dollars. Further information about flows can be gleaned from statistics concerning nonfinancial cross-border transactions. Mexican exports bring dollars to Mexico, and Mexican imports take them away. In this paper, total flows of dollars into Mexico are measured by interbank shipments plus a constant multiple of Mexican border income from exports of goods and services to the United States; dollars leaving Mexico are measured by dollar receipts of U.S. banks from Mexican banks plus a constant multiple of Mexican border expenditures on imports of goods and services from the United States. There are, in addition, other sources of net flows, which, for lack of data, are assumed proportional to income. The estimated equation, using quarterly data from the first quarter of 1973 through the fourth quarter of 1981, is 6.22 - 1.73 (.6) (- 4.5)  It  Y + t  1.67 (3.6)  X  Yt t  2.39 ( - .3)  tJ.Y  Yt t  The coefficients of the fourth and fifth variables were constrained to equal the coefficients of the last two variables. The figures in parentheses are t statistics; R2 = .76, DW = 1.37, and rho = .65. The variables are defined as follows: 5 = net shipments of U.S. currency to Mexican banks from Eleventh District banks (Millions of dollars, at annual rates) Y  =  estimated GDP of the four Mexican states that border on the Eleventh District 8  = Mexican income from border transactions (Millions of dollars, at annual rates)9 Economic ReviewlJuly 1983  x=  Mexican expenditures in border transactions (Millions of dollars, at annual rates)9  E  a measure of exchange rate risk (defined in the text below)  = rate on 90-day dollar certificates of deposit (CDs) t = time.  f  The coefficients fort and are significant and correctly signed. They indicate the importance of nonfinancial border transactions in explaining currency shipments. The next three variables in the equation attempt to capture the currency shipments that result simply from income changes in the border states. An increase in income should raise the demands for both currencies without causing any currency substitution. Collectively, the estimated parameters provide weak support for that supposition. While the coefficient of the first of these variables has the wrong sign, it is statistically insignificant. And because the average value of is about 2, the total estimated effect of an increase in income is generally positive.  +  8. Estimates for quarterly income for the border area of Mexico adjacent to the Eleventh District were constructed as follows. Annual gross domestic product (GOP) for all of Mexico was multiplied by the annual ratio of electricity generated in the four states bordering on the Eleventh District (Chihuahua, Coahuila, Nuevo Leon, and Tamaulipas) to electricity generated in the whole country. This measure of annual GOP for the four-state area was interpolated with a quarterly series of Mexican industrial production to provide an estimate of quarterly GOP for the four states. The interpolation was performed by the Chow-Lin procedure using the first-order Markov process errors, as described in Gregory C. Chow and An-Ioh Lin, "Best Linear Unbiased Interpolation, Distribution, and Extrapolation of Time Series by Related Series," Review of Economics and Statistics 53 (November 1971): 372-75. Thanks are due to Robert Litterman for programming assistance. 9. Mexican income from border transactions (transacciones fronterizas- ingresos) and Mexican expenditures in border transactions (transacciones fronterizas-egresos) are data compiled from surveys by the Mexican government. The data include information from businesses on sales of merchandise (for income) and purchases of merchandise (for expenditures) for purposes of both investment and current consumption. The data also include information on purchases of merchandise and services by individuals (for the expenditure portion), including tourist expenditures, and (for the income portion) include professional fees and other income. The survey for business and other institutions is different from the survey for individuals.  17  The last two variables capture the currency substitution effect. Using a measure of exchange rate risk in order to determine whether currency substitution is significant has been done before, although not in an equation that examines currency flows.lO The futures market premium or discount can be used to calculate an average expected change in the spot rate. The expected percentage change in the spot rate is calculated as follows. Let F equal the futures rate and R equal the spot rate as previously defined. " Then  E = 100(F-R)/R. In a fixed-rate or closely managed floating-rate regime such as Mexico's, the "expected" change is the average of possible rate changes, including no change at all, weighted by their probabilities of occurring. The ratio of the opportunity cost of holding peso currency (as compared with holding interest-earning dollar deposits) to the opportunity cost of holding dollar currency is equal to (i + E)/ i or 1 + (E Ii). The ratio of E to i can be thought of as the proportion by which exchange risk raises the cost of holding pesos relative to dollars. The estimated coefficients of the exchange rate variables indicate that currency shipments (and, by implication, dollar usage in northeastern Mexico) are substantially affected by changes in the peso's perceived exchange risk or expected future change in value. With dollar CD rates at 10 percent, an increase in the 90-day future peso/dollar rate amounting to 15 percent of the spot rate would raise U.S. currency demand across the border from the Eleventh District by about a quarter of a percent of annual income in that four-state area, or over $100 million at the late-1981 income level.  Conclusion The Mexican exchange controls imposed in 1982, the effective abolition of dollar-denominated bank accounts in Mexico, and other policies of the Mexican government hindered the ability of Mexican households and businesses to diversify their currency portfol ios. I n the wake of 1982's real ignment of exchange relationships between Mexico and the United States, interbank currency shipments between the Eleventh District and Mexico were greatly attenuated. These movements continue to occur only at very low levels. For example, in the second 18  quarter of 1981, gross shipments of dollars to Mexico by Eleventh District banks reached $143.2 million, while receipts of dollars from Mexican banks totaled $51.8 million. In the second quarter of 1983, gross shipments were only $2.8 million while gross receipts were $31.6 million. Thus, the value of net shipments of dollars declined from + $91.4 million to -$28.9 million. Even so, discussions and expressions of concern about dollarization continue to appear in the Mexican press. 12 Such concern is not unwarranted. Common statistical measures of movement by Mexicans out of pesos and into dollars suggest that currency substitution has declined. Institutional changes have been implemented with the intent of reducing such substitution. The need of Mexicans to reduce exchange rate risk has led to increased use of techniques for the surreptitious capture of dollar reserves in the United States, such as overinvoicing of imports and underinvoicing of exports. Continuing long-term differentials between U.S. and Mexican inflation rates in the face of an insufficiently adjusting peso may offer the incentive for currency substitution's eventual resurrection within Mexico. If history is any guide to the future, once sufficient motivation arises for currency substitution, institutional impediments will somehow be circumvented. It has been claimed that Mexico's northern border can almost be considered a currency area separate  10. Bordo and Choudhri, cited in footnote 4, test for the existence of currency substitution throughout Canada with a money demand function, using a measure of expected change in the spot rate as an argument in their function.  11. The spot rate for pesos is the pesos-per-dollar exchange rate in current trading for current delivery in the open market. The futures rate for pesos is expressed in pesos per dollar at a price established for delivery of pesos three months in the future. as quoted in the International Monetary Market of the Chicago Mercantile Exchange. Observations for a given quarter are the average prices in the last month of a given quarter for pesos to be delivered in the last month of the next quarter. Thus. for the first quarter of a given year, the futures price is the average price in March for delivery in the month of June. This procedure was used because there are only four delivery months in any year in the International Monetary Market. Each delivery month is the last month in the quarter. Thus. the delivery months for a given year are March. June, September. and December. 12. See. for example. Abel Beltran del Rio, "Tasa de inten?s e inflacion." Transformacion-. May 1983.11-17.  Federal Reserve Bank of Dallas  from the rest of Mexico. Because of the considerable evidence of the special quality of the border, the test performed in this research cannot be said to examine the existence of currency substitution with regard to bills and coinage throughout Mexico. However, it does offer evidence of changes in Mexican border demand for U.S. dollars to an extent that cannot be explained simply by changes in purchases of U.S. goods and services or through changes in income. The evidence suggests that during portions of the 1970's and early 1980's, variables that may be associated with currency substitution also had additional, statistically significant explanatory power for changes in interbank currency shipments and, presumably, for changes in the overall demand for dollar-denominated currency in the four border states of Chihuahua, Coahuila, Nuevo Leon, and Tamaulipas. Although dollarization, as observed in dollardenominated Mexican bank deposits, has occurred for more than 50 years, it declined in the 1960's and  very early 1970's. During this period, Mexico's inflation rate was not greatly different from that of the United States, so the incentives to substitute dollars for pesos were not very powerful. The more recent instability in the Mexican economy was preceded by many years of economic stability and gradual growth. Currency substitution, to the extent that it limits a country's ability to pursue a policy greatly more expansive than policies of its chief economic partners, may serve to encourage a return to more stable economic management. Recent Mexican experience indicates that a great deal of pol icy divergence may occur even with currency substitution, and this divergence cannot continue forever even without it. However, developments in Mexico also may serve as examples of how currency substitution can ultimately help restrain a country from its excesses. Currency substitution can clearly become one of the consequences of such excesses and may be one of the forces leading in the long run to exchange rate adjustment.  Appendix Derivation of the Estimating Equation and Identification of Parameters Assume dollar and peso currencies provide monetary services according to the following relation:  o < a,(3,b < 1; A > 0, = o =  where M  Border residents and businesses maximize the flow of monetary services less the opportunity costs of not holding dollar-denominated time deposits. The opportunity costs are  iD where  flow of monetary services dollar currency in the border area of Mexico, adjacent to the Eleventh Federal Reserve District  P  =  dollar value of peso currency in the same border area of Mexico  Y  =  dollar value of income in the same border area of Mexico.  Economic Review/July 1983  +  (i+E)P,  i = yield on 90-day dollar deposits divided by 4 E  = (90-day future peso/dollar exchange rate less the spot rate) divided by the spot rate.  From the first-order conditions:  ~ = (ba  .  ij Et(1-a)  ::::: (ba)1/(1-a)  +  E (ba)1/(1-a) 1-a .  T. 19  Assume that  1:.0  and  = (S-R) +  +  ff +gX  g<O<f.  hY,  where 5 = currency shipments to Mexico  R  =  currency receipts from Mexico  f = dollar value of receipts from exports to the United States (ingresos)  The parameter restrictions are  X = dollar value of imports to the United States (egresos). If the full response to changing expectations about the exchange rate is not instantaneous, the equation can be supplemented with  Then,  1:.0:::: (bO')1/(1-a) I:.Y  +  (bO')l/(l - a) (  1-0'  E  -:-1:. Y I  +  E)  Y 1:."7 I  '  and  (S-R)  Y  -h - f ~  +  - g~  +  (bO')l/(l-a)(i. I:.Y  1-0'  i  (bO')l/(l-a) .  +  Y  I:.f)  i'  t:.;  where the -1 subscript denotes a one-period lag to allow lagged adjustment. In practice, the estimate of a3 was too imprecise to allow computation of the structural parameters.  The estimating equation is  E Y I:.Y + a4 (T' + I:. TE) ,  20  Federal Reserve Bank of Dallas  FEDERAL RESERVE BANK OF DALLAS STATION K, DALLAS, TEXAS 75222 ADDRESS CORRECTION REQUESTED  BULK RATE U.S. POSTAGE  PAID PERMIT NO. 151  The Economic Review is published by the Federal Reserve Bank of Dallas and will be issued six times in 1983 (january, March, May, July, September, and November). This publication is sent to the mailing list without charge, and additional copies of most issues are available from the Publications Department, Federal Reserve Bank of Dallas, Station K, Dallas, Texas 75222. Articles may be reprinted on the condition that the source is credited. The views expressed are those of the individual authors and do not necessarily reflect official positions of the Federal Reserve Bank of Dallas or the Federal Reserve System.