Domestic Competition, Cyclical Fluctuations, and Long-Run Growth in Hong Kong Sar
Author: Mr. Eric Zitzewitz1
  • 1 0000000404811396https://isni.org/isni/0000000404811396International Monetary Fund

Contributor Notes

Author’s E-Mail Address: ericz@mit.edu
Publication Date:
01 Aug 2000
eISBN:
9781451856248
ISBN:
9781451856248
Language:
English
Keywords:
WP; Sar; P-AC margin; P-MC; P-AC markup; Hong Kong SAR-OECD sample; P-MC markup; markup cyclicality; value-added P-AC margin
Download PDF (1,021.2 KB)

This paper provides an empirical assessment of the degree of competition in Hong Kong SAR using industry-level data. Although due to data limitations only approximate measures of competitiveness can be estimated, the results do suggest that Hong Kong SAR is as competitive as a typical OECD economy. The dramatic shift of the economy toward services over the last decade has also made it slightly less competitive on average. Imperfect competition is not leading to counter-cyclical markups and slower price adjustment as some theories predict, however, since markups are more pro-cyclical than in OECD countries. Lastly, markups are sufficiently imperfectly competitive in both Hong Kong SAR and the OECD to significantly downwardly bias growth accounting estimates of total factor productivity in Asian NICs vis-à-vis OECD countries.

Abstract

This paper provides an empirical assessment of the degree of competition in Hong Kong SAR using industry-level data. Although due to data limitations only approximate measures of competitiveness can be estimated, the results do suggest that Hong Kong SAR is as competitive as a typical OECD economy. The dramatic shift of the economy toward services over the last decade has also made it slightly less competitive on average. Imperfect competition is not leading to counter-cyclical markups and slower price adjustment as some theories predict, however, since markups are more pro-cyclical than in OECD countries. Lastly, markups are sufficiently imperfectly competitive in both Hong Kong SAR and the OECD to significantly downwardly bias growth accounting estimates of total factor productivity in Asian NICs vis-à-vis OECD countries.

I. Introduction

Hong Kong SAR has traditionally been viewed as having highly competitive product and factor markets. Competition has been credited not only with contributing to high long-run output and productivity growth but also with providing a supporting framework for rapid price adjustment in response to economic shocks while maintaining a fixed exchange rate. Given the importance of competition, surprisingly little systematic empirical work has been done in evaluating the degree of competition in Hong Kong SAR’s industries. While prices have historically adjusted much more rapidly in response to shocks in Hong Kong SAR than in OECD economies (Box 1), many observers in recent years have raised questions about the degree of domestic competition, particularly in light of the economy becoming increasingly dominated by service industries.2

This chapter estimates quantitative, industry-level measures of the intensity of competition and compares them both across industries and between Hong Kong SAR and OECD countries. These measures are also used to evaluate whether degree of competition has changed in the last decade and how markup cyclicality affects the adjustment process. The chapter has four main findings:

  • Hong Kong SAR is neither significantly more nor significantly less competitive than the average OECD country. Given its fixed exchange rate and exposure to external shocks, however, Hong Kong SAR may actually need more competition than other countries. Further investigation of competition in specific sectors may therefore be justified. Most of the measures in this paper suggest that food processing, nonmetallic minerals, real estate and transportation are less competitive than the average industry, while construction, restaurants and hotels, and textiles and apparel appear to be among the most competitive.

  • Competition in Hong Kong SAR has become slightly less intense over the last decade as the economy has rapidly shifted out of manufacturing.

  • Markups are more pro-cyclical in Hong Kong SAR than in the average OECD country in the service sector, which now accounts for over 90 percent of GDP. The shift to services has also made economy-average markups more pro-cyclical. This should be contributing to faster-than-average price adjustment in Hong Kong SAR.

Price Adjustment in Hong Kong SAR and OECD Countries

The rapid adjustment of prices in Hong Kong SAR in the past is illustrated by the slope of the Phillips Curve, i.e., the effect of the level of the unemployment rate on the acceleration of the inflation rate. A single point increase in the unemployment rate has historically been associated with a deceleration of 0.41 points in inflation in Hong Kong SAR, compared with 0.05 points for the average country in our OECD sample. The conclusion that price adjustment has been extremely rapid in Hong Kong SAR is robust to controlling for external price shocks (represented by the acceleration of price inflation for imports, exports, or oil prices) and unemployment hysterisis effects (handled by including lagged unemployment in the regression equation).

Phillips Curve in Hong Kong SAR and OECD, 1973–98 Dependent Variable: Change in Year-on-Year Rate of CPI Inflation

article image
Note: Regressions use quarterly data. Lagged unemployment is lagged by 4 quarters. Regressions are estimated using OLS and are not intended to be a prediction of the effect of unemployment on the acceleration of inflation, but rather a measure of their correlation and relative variance. Standard errors are in parentheses.

  • Market imperfections in both Hong Kong SAR and the OECD cause labor shares from national accounts data to understate the contribution of labor in output growth. Using an underestimated labor share downwardly biases measured total factor productivity (TFP) growth for countries with rising capital-labor ratios, for example, the Asian NIEs. Correcting this bias increases the contribution of TFP to growth in Hong Kong SAR. Moreover, such corrections to the growth accounting methodology could increase the productivity growth of all Asian Newly Industrialized Economies (NIEs) relative to OECD countries.

The remainder of the chapter is organized as follows. The next section discusses various methodologies and data used for measuring competition. The data in Hong Kong SAR is limited along two key dimensions, and thus several imperfect measures of competition are experimented with. The third section compares the level of the intensity of competition in Hong Kong SAR with that of OECD, while the fourth section examines how the intensity of competition in Hong Kong SAR has changed over time. The fifth section compares the cyclicality of markups in Hong Kong SAR and the OECD and discusses the implications for the adjustment process. A final section examines the implications of these findings about markups for the measurement of total factor productivity and thus for long-run growth in Hong Kong SAR.

II. Methodology and Data

A. Some Basic Concepts

Under Cournot competition, a profit maximizing firm sets its quantity such that:

(P−MC)/P=(1/εd)/neffP=εd⋅neff/(εd⋅neff−1)⋅MC

where P is the price of the good, MC is its marginal cost, εd is the market elasticity of demand and neff is the effective number of firms in the market, which is equal to one if firms are perfectly colluding and the actual number of firms if they are behaving noncooperatively. Note that higher markups do not necessarily imply less competition; they may also be due to lower demand elasticity. If it is assumed that the elasticity of demand for the output of a given sector is constant across countries, then a cross-country comparison of markups can reveal differences in competition. Comparisons of markups across sectors within a country can be more problematic, however, since demand elasticities can be quite different across sectors.

There are several theories that predict that markups will vary with the business cycle. In the two most popular—Rotemburg and Saloner (1986) and Green and Porter (1984)—markups vary because the ability of firms to collude, and thus neff, varies with the business cycle. In the Rotemburg and Saloner model, collusion is harder to sustain in booms because the benefits of cheating are proportional to current demand but the benefits of maintaining collusion are proportional to long-run demand. Sustainable markups are therefore counter-cyclical.3 In Green and Porter, collusion is harder to sustain in recessions since firms misinterpret a fall in market demand as cheating and this leads to price wars and thus pro-cyclical markups, causing average markups to be pro-cyclical.4 Counter-cyclical markups appear to be more common empirically, at least in the United States.5

In order to measure the cyclicality of markups, consider the following relationship between price-marginal cost (P-MC) and price-average cost (P-AC) markups (Morrison, 1990):

P/AC=(CU/RTS)⋅(P/MC)

where CU is the capacity utilization of an industry (defined as the long-run minimum cost to produce the current quantity divided by the current total cost) and RTS is the long-run return to scale (defined as the ratio of marginal cost to minimum average cost). Since capacity utilization is usually strongly procyclical and returns to scale are roughly constant through the business cycle, P-AC markups can be pro-cyclical even when P-MC markups are countercyclical. This is in fact what the literature has found for the United States.6 If the variation of capacity utilization and returns to scale with the business cycle is similar across countries within a given sector, then more procyclical P-AC markups are in country, then less countercyclical are P-MC margins.7 Likewise, if the average level of capacity utilization and returns to scale are similar across countries, comparisons of the level of P-AC markups can be revealing about P-MC markups.

B. Estimation Techniques

Three broad approaches have been taken to estimating P-AC and P-MC markups and margins:8 direct measurement of P-AC margins, estimation of marginal cost by instrumenting for output, and estimating P/MC, CU, and RTS using a structural model (Morrison, 1990). In this paper, both of the first two methods will be used. The third method is more desirable, but is impossible for Hong Kong SAR given the data available.9

  • Direct measurement of P-AC. Domowitz, Hubbard, and Petersen (DHP) (1986) directly measured price and average variable cost using census data. This approach is the most straightforward and allows one to calculate an independent measure of the P-AC margin in each year, allowing one to estimate margin cyclicality even with short time series. Drawbacks of this approach include the difficulty of measuring the economic cost of capital services and the fact that, as discussed above, P-AC margins are more difficult to interpret as measures of competition than P-MC margins.

Since time series are too short in Hong Kong SAR to allow the construction of capital stocks, P-AC margins will be measured without any allowance for capital costs. DHP (1986) noted that under the assumptions that labor is completely variable, capital is completely fixed, and short-run returns to scale are constant, a P-AC margin which excludes capital costs is equal to a P-MC margin. To the extent that these assumptions are violated, the P-AC margins measured in this paper will be imperfect measures of P-MC margins.

  • Econometric estimation of P-MC margins. In this paper three methodologies for estimating P-MC margins were considered: Hall (1988), DHP (1988), and Roeger (1995). All of these methods assume a constant margin over time, constant returns and no fixed or adjustment costs. Only a modified version of the Roeger methodology is feasible for Hong Kong SAR, but this “modified Roeger” method yields results which are highly correlated with those from other methods.

The Hall and DHP methodologies both use instrumental variables to identify changes in an industry’s output which are not related to changes in its productivity.10 These exogenous changes in output are then used to infer marginal cost and P-MC margins. The respective Hall and DHP estimating equations are:

dq−dk=μ⋅a⋅(dl−dk)+d(lnA)dTFP=γ⋅(dq−dk)+(1−γ)⋅d(lnA)

where dq, dk, dl, and dTFP are changes in the logs of output, capital, labor, and measured TFP, a is the observed labor share, μ is the P-MC markup, and γ is the P-MC margin. 11 The change in unobserved true productivity (dlnA) is treated as the error term in the estimation, and the identifying assumption is that the demand instrumental variables are uncorrelated with true productivity growth. Hall used defense spending, world oil prices, and the political party of the President as demand instruments for U.S. industries, but these instruments have been criticized as being only weakly correlated with output and inappropriate for other countries. Fortunately, DHP found that using aggregate GDP as a demand instrument yielded similar results, and this is the instrumenting approach followed in this study.12

The Roeger method, in contrast, eliminates the unobserved productivity term from the estimating equation, removing the need for instrumental variables.13 Instead of using the primal (output-based) Solow residual, Roeger takes the difference between the primal and dual (price-based) residuals, eliminating the productivity term. His estimating equation is:

(dq+dp)-a⋅(dl+dw)−(1-a)⋅(dk+dr)=γ⋅[(dq+dp)−(dk+dr)]

Notice that the Roeger method requires only nominal data, which is very helpful given the lack of industry-level price deflators in Hong Kong SAR. It does require data on current price capital inputs, however. Since capital stocks are unavailable for Hong Kong SAR, a modified Roeger methodology using gross fixed capital formation as the capital input is estimated.14

C. Data Availability and Sources

Data is available for Hong Kong SAR for 85 industries (26 manufacturing and 59 service) that cover 5 out of 9 major sectors in the economy.15 As mentioned above, the available data has several important limitations, however. First, only 8 years (1990–97) and 12 years (1986–97) of data are available for manufacturing and service industries, respectively.16 This makes it impossible to construct industry-level capital stocks and harder to identify marginal costs and the cyclicality of markups. The latter is complicated by the fact that the particular 12 years for which we have data was one of the most stable periods in Hong Kong SAR history, especially the 8 years (1990–97) for which we have manufacturing data. Second, industry-level output and input price deflators are not available for Hong Kong SAR. These deflators are especially important for Hong Kong SAR given the dramatic changes in relative prices in during the period studied,17 and their absence restricts us to using the Roeger methodology for estimating P-MC margins.18

The Hong Kong SAR data is matched with data for 14 countries19 from the OECD ISDB and STAN databases. The OECD data is available for 27 manufacturing industries and 11 nonmanufacturing industries; 26 and 7 of which match with the Hong Kong SAR data. Due to the shortness of the manufacturing time series, the 26 matching manufacturing industries are pooled into the 14 manufacturing categories in the ISDB, which helps increase the power of estimates for Hong Kong SAR. Value-added data, and thus margins, are unavailable for communication and finance/insurance for Hong Kong SAR, and thus the combined Hong Kong SAR-OECD sample includes 14 manufacturing and 5 nonmanufacturing industries.

III. How Competitive is Hong Kong SAR?

The three measures of margins presented in this section yield slightly inconsistent conclusions about how the competition of Hong Kong SAR’s industries compares with the OECD or the United States (Figure 1):

  • Price-average cost margins defined on value added are generally higher in Hong Kong SAR, especially in manufacturing (Table 1).

  • P-AC margins defined on gross output are slightly higher on average in Hong Kong SAR in manufacturing, and slightly lower in services (Table 1). 20

  • P-MC margins, as estimated by the modified Roeger methodology, are lower in Hong Kong SAR in both manufacturing and services (Table 2).

Figure 1.
Figure 1.

Industry-by-Industry Comparison of Intensity of Competition in Hong Kong SAR Versus the OECD

Citation: IMF Working Papers 2000, 142; 10.5089/9781451856248.001.A001

Sources: Hong Kong Census and Statistical Department; and staff estimates.Notes:1. Much more competitive is defined as a 12 percent lower value added P-AC margin, a 4 percent lower gross output P-AC margin, or a significantly lower OLS P-MC margin. More competitive is defined as a 3 percent lower value added P-AC margin, a 1 percent lower gross output P-AC margin, or an insignificant difference of-.03 in the OLS estimate of the P-MC margin.2. Much more pro-cyclical is defined as a significantly higher cyclicality coefficient, more pro-cyclical is defined as an insignificant difference in cyclicality greater than 1.0.3. Much less and less competitive/pro-cyclical are the reverse of these definitions. Significant means significant at the 5 percent level.
Table 1.

Hong Kong SAR: Price-Average Cost Margins, 1986–97 Average

(In percent)

article image
Sources: Hong Kong Census and Statistical Department; and staff estimates.
Table 2.

Hong Kong SAR: OLS and IV Estimates of P-MC Markups Using the Modified Roeger Methodology

article image
Sources: Hong Kong Census and Statistical Department; and staff estimates.Notes:1. Regressions arc pooled for industries containing more than one component industry.2. Significance at the 10, 5, and 1 percent level is indicated by one, two, and three asterisks, respectively.

Although comparing Hong Kong SAR with the OECD average yields different results depending on the measure, on almost all of measures, Hong Kong SAR is neither the most nor the least competitive of our 15 sample countries (Table 3). This suggests that Hong Kong SAR is neither significantly more nor significantly less competitive than OECD countries.

Table 3.

Hong Kong SAR’s Relative Rank on Different Measures of Competitiveness

article image
Sources: Hong Kong Census and Statistical Department; and staff estimates.Notes:1. A lower rank number implies more competition or more pro-cyclical margins.2. Rankings are out of the 15 sample countries, except for P-MC margins, where they are out of 13.

Of course, it may be the case that given its fixed exchange rate and extreme openness and exposure to external shocks, Hong Kong SAR needs to be significantly more competitive than the average OECD economy. Figure 2 compares the results in this section for Hong Kong SAR across industries to identify industries in which competition could potentially be stronger. These comparisons require stronger assumptions than comparisons across countries, as discussed in Section II, but may nonetheless be informative. The plurality of the measures in Figure 2 suggest that food processing, nonmetallic minerals, real estate and transportation are less competitive than the average industry, while construction, restaurants and hotels, and textiles and apparel are among the most competitive. It should be noted, however, that the nonmetallic minerals industry comprises an extremely small portion of total manufacturing output.

Figure 2.
Figure 2.

Comparison of Intensity of Competition Across Industries in Hong Kong SAR

Citation: IMF Working Papers 2000, 142; 10.5089/9781451856248.001.A001

Sources: Hong Kong Census and Statistical Department; and staff estimates.

Potential Biases in the Competition Measures

Both the P-AC and the estimated P-MC margins are imperfect estimates of true P-MC margins, as discussed in Section II. Recall that for P-AC margins that exclude capital costs to equal P-MC margins, capital would have to be completely fixed, labor completely variable, and returns to scale would have to be constant. Obviously these assumptions do not hold in reality, but the primary concern in this paper is whether making them distorts the comparison between Hong Kong SAR and the OECD. For example, even if capital and labor were both only partly fixed, if they were fixed to the same degree and if factor shares were similar in Hong Kong SAR and the OECD, the Hong Kong SAR-OECD difference in P-AC margins should still be an unbiased indicator of differences in P-MC margins.

On the other hand, if Hong Kong SAR uses technologies or has a product mix that involved less fixed costs across all factors (and thus higher MC/AC ratios), it could have simultaneously higher P-AC margins and lower P-MC margins. Furthermore, if Hong Kong SAR has a higher capital share in value-added or output, this could upwardly bias its measured P-AC margin differentially. 21 Given the limited data available for Hong Kong SAR, there is little one can do but accept that the P-AC margins may be slightly biased; this is part of why this paper contains multiple measures.

The P-MC margins estimated by the “modified Roeger” methodology are likewise imperfect measures. The IV results for Hong Kong SAR are fairly erratic with high standard errors, thus we rely mainly on the OLS results (Table 2). Comparing the modified Roeger method with the DHP and unmodified Roeger methods for U.S. and OECD data suggest that the OLS modified Roeger estimates may be downwardly biased (Table 4). This would be the case if growth in gross fixed capital formation was either an especially noisy proxy for growth in current price capital input.22 As with P-AC margins, we are especially concerned with sources of differential bias. The period studied was actually a fairly stable one for Hong Kong SAR; from 1986-97 the standard deviation of the growth rate in aggregate GFCF was 4.3 percent in both Hong Kong SAR and the United States, suggesting that any bias due to measurement error should be comparable across at least these two countries. For OECD countries, the OLS version of the modified Roeger method yields results that are highly correlated with other methods across both countries and industries, so the slightly lower P-MC margins estimated for Hong Kong SAR by the modified Roeger methodology suggest that other methodologies would also have produced slightly lower estimates (Table 5).

Table 4.

Hong Kong SAR: Four Methodologies for Measuring P-MC Margins

article image
Sources: Hong Kong Census and Statistical Department; and staff estimates.Notes:1. The number below the OLS-IV difference in italics is the p-value of a Hausman(1978) test.2. The U.S. dataset used in these tests is the Bartlesman-Becker-Gray (1996) dataset collapsed down to the 2-digit SIC level, which should be very similar to the data used by Hall (1988) and Roeger (1995).
Table 5.

Hong Kong SAR: Correlation of P-MC Margin Estimates Using Different Methodologies

article image
Sources: Hong Kong Census and Statistical Department; and staff estimates.Notes:1. Bold italics indicate correlation coefficients highlighted in the text.2. Although the Hall methodology estimates P/MC and the others estimate (P - MC)/P, the correlation should still be high for the relevant range of values.

IV. Evolution of Competition in the Hong Kong SAR Economy

This section examines both whether margins have increased or decreased within industries and whether the rapid sectoral shift which has occurred in Hong Kong SAR has shifted the economy into less competitive industries. As with margin levels, the different measures are inconsistent. The balance of evidence, however, suggests that Hong Kong SAR has become slightly less competitive in the last 10 years:

  • Value-added P-AC margins have declined in services, but they have risen in manufacturing and gross output P-AC margins have risen in both sectors (Table 6).23

  • P-MC margins, as estimated by the modified Roeger methodology, have been stable in services and declining slightly in manufacturing, although this decline is not statistically significant.24

  • The shift in the economy towards services has raised average P-MC margins while having a minimal effect on average P-AC margins (Table 7).25

Table 6.

Hong Kong SAR: Trend in Price-Average Cost Margins

(Percentage change per year)

article image
Sources: Hong Kong Census and Statistical Department; and staff estimates.Notes:1. One, two, and three asterisks represent significance at the 10 percent, 5 percent, and 1 percent level, respectively.2. Numbers reported are coefficients on the time trend in a regression of the P-AC margin on a time trend and the GDP gap (for the cyclically adjusted measure).3. The GDP gap is constructed by applying a Hodrik-Prescott filter (lamba = 1600) to quarterly data and then aggregating the quarterly gap into annual data.
Table 7.

Hong Kong SAR: Effect of Industry Mix Shift on Margin Levels and Cyclicality

(In percent)

article image
Sources: Hong Kong Census and Statistical Department; and staff estimates.Note: Higher numbers imply less competition, except for P-AC margin cyclicality.

The magnitudes of some of the changes in margins are quite large relative to the Hong Kong SAR-OECD level differences discussed in the previous section. Table 6 suggests that value-added margins have increased by 7.1 percentage points over 10 years in manufacturing and decreased by 6.0 percentage points in services, while gross output margins have increased by 1.9 and 1.4 percentage points, respectively. Table 7 shows that the shift in industry mix toward services should have increased the economy-average P-MC margin by 7.7 percentage points. 26 In contrast, the Hong Kong SAR-OECD level differences in value-added P-AC, gross output P-AC, and P-MC margins were approximately 10, 1, and 2 percentage points, respectively. Given the inconsistency of the measures and their potential biases (discussed in Section III), one should be cautious in drawing firm conclusions above the change in the level of competition in Hong Kong SAR.

V. Cyclicality of Markups and Adjustment in Hong Kong SAR

One of the objectives of this chapter was to understand whether imperfect competition has adversely affected cyclical adjustment in Hong Kong SAR. If imperfect competition generates countercyclical markups, as predicted by the Rotemburg-Saloner (1986) model, a worsening of competition could slow down price adjustment, creating problems for an open economy with a fixed exchange rate.

An Illustration of How Cyclicality in Markups Affects Adjustment

To see how imperfect competition can affect the adjustment process, consider a simple model of aggregate supply and demand. Expenditure, and thus aggregate demand, is determined by a simple quantity theory of money:

PY=VM

Where Y refers to aggregate output, V is the velocity of money, and M is the stock of money supply. Prices are set by firms which mark up their marginal wage (W) and material costs (MAT) depending on the degree of competition in their markets:

P=c(MAT,W)⋅MU

The aggregate supply relationship governs how sticky price adjustment in response to deviations of aggregate output from its long-run trend is:

dp(y)=sw⋅dw(y)+sm⋅dmat(y)+dmu(y)

where lower case letters indicate logarithms, y is the deviation of output from its long-run level and sw and sm are the share of wages and materials in cost. Ceteris paribus, if markups are more countercyclical, prices will fall less rapidly during recessions and rise less rapidly during booms.

Consider the effects of an adverse monetary shock to aggregate demand. Prices are initially sticky, so the immediate effect of the monetary shock is an equally sized decline in both nominal and real output. A below trend real output level, however, leads to a decline in prices and thus a recovery of real output which continues until output reaches its long-run level and the effect of the monetary shock is fully reflected in prices. If markups are countercyclical, price adjustment is slower and the recession lasts longer.

Recessions resulting from a supply shock are also more severe when markups are countercyclical. Consider the effects of a permanent increase in oil price inflation. If monetary poUcy is unchanged, the increased oil price inflation will lead over time to a higher price level and lower real output level. This process continues until real output has fallen enough so that changes in wages and markups are reduced enough to restore price stability (or, alternatively, until inflation returns to equal the rate of money creation). Absent a change in monetary policy or a reduction in oil price inflation, real output is permanently lower. The size of the decline in real output is greater when markups are counter-cyclical, since more real output decline is required to offset the higher oil price inflation.

How Cyclical Are Markups in Hong Kong SAR?

This section measures the cyclicality of P-AC margins in Hong Kong SAR compared with OECD countries and tests whether the structural shift towards services has made P-AC markups more counter-cyclical. As discussed in Section II, the relative cyclicality of P-MC margins can be inferred from the cyclicality of P-AC margins if we assume that capacity utilization responds identically to industry output in all countries. 27 Contrary to expectations, it was found that:

  • P-AC margins are significantly more pro-cyclical in Hong Kong SAR than in the OECD in services, while they are slightly more counter-cyclical in manufacturing (Table 8).

  • P-AC margins are more pro-cyclical in services. The shift in the economy to service therefore should have made aggregate P-AC margins more pro-cyclical (Table 7).

Table 8.

Hong Kong SAR: Cyclicality of Price-Average Cost Margins

(Percentage point effect on the margin of a one percentage point change in the output gap)

article image
Sources: Hong Kong Census and Statistical Department; and staff estimates.Notes:1. One, two, and three asterisks represent significance at the 10%, 5%, and 1% level, respectively.2. Numbers reported are coefficients on the GDP gap in a regression of the P-AC margin on the GDP gap and a time trend. Regressions are for the OECD countries and for Hong Kong SAR industries containing more than one component industry.3. The GDP gap is constructed by applying a Hodrik-Prescott filter (lamba = 1600) to quarterly data and then aggregating the quarterly gap into annual data.

Therefore there is little evidence that increased counter-cyclicality of margins is slowing price adjustment in Hong Kong SAR. If one concludes that price adjustment has slowed, one may have to look elsewhere (e.g., factor markets) for the explanation.

Can we infer anything about competition from the cyclicality of margins? A majority of studies (at least as summarized in Rotemburg and Woodford, 1999) suggest that imperfect competition has created counter-cyclical P-MC margins, at least in the United States. If it is also true that more collusion means more counter-cyclical markups, which is what the Rotemburg-Saloner (1986) model predicts at all but very high levels of collusion, then the cyclicality of P-AC margins in Hong Kong SAR imply that services are more competitive and manufacturing is less competitive. To feel comfortable with this conclusion, however, one would have to eliminate all other sources of differences in cyclicality, such as differences in technology or the cyclical behavior of demand elasticity, which is difficult to do given the data available.

VI. Implications for the Measurement of TFP Growth

As Hall (1990) and Solow (1957) have observed, the traditional Solow-residual measure of TFP using observed factor shares is valid only under perfect competition. Under imperfect competition, the observed labor share will be below the coefficient in the Cobb-Douglas production function. If labor is 100 percent variable, the observed labor share will be:

a=WL/PQ=α/MU

where a is the observed labor share, α is the production function coefficient, and MU = P/MC. Assuming constant returns to scale, measured two-factor TFP is equal to:

lnTFP=a⋅(MU−1)⋅(lnl−lnk)+lnA

where TFP is the Solow residual and A is true productivity. 28 The source of the difference between measured and true productivity is the difference between the observed and true labor share when there is market power. Underestimating the labor share creates a downward bias in measured TFP growth in economies with rising capital-labor ratios, such as the Asian NIEs. 29 Table 9 shows the magnitude of this bias for both OECD countries and the Asian NIEs assuming a average markup in every country of 1.5, one of the lower estimates from this paper.30 Estimates are shown for both the case where labor is 100 percent variable and where it is 80 percent variable.31

Table 9.

Hong Kong SAR: Bias to Measured TFP Implied by the Estimated Markups

(Average annual growth rates)

article image
Sources: Hong Kong Census and Statistical Department; and staff estimates.Notes:1. Data for the Asian NICs are from Young (1995), data for the G4 are from the OECD Economic Outlook. Labor and capital inputs are Young’s adjusted inputs for the NICs and total employment and business-sector capital stock for the G4.2. Bias is calculated assuming an economy-wide markup of 1.5 in every country.

Assuming labor is 100 percent variable, correcting for the imperfect competition bias raises measured TFP growth by about 0.3 percent a year for the United States, Germany, and France and 1.5 percent to 2.5 percent a year for the Asian NIEs. The adjusted labor shares are clearly too high to be realistic, however. Assuming that labor is 80 percent variable lowers the TFP adjustment to 0.5 percent to 1.0 percent for the NIEs and 0.1 percent to 0.2 percent for the advanced economies and produces at least plausible labor shares. The main conclusion to be drawn from this analysis is that comparisons of TFP growth between rapidly industrializing and advanced economies are sensitive to the factor shares used, the correct measurement of which is in turn sensitive to assumptions about imperfect competition and factor fixity. 32 These issues are not as likely to affect comparisons between rapidly industrializing countries whose capital-labor ratios are changing at roughly the same rate. From 1966–90, Hong Kong SAR achieved growth which was 1.2 percent p.a. slower than in Singapore despite capital and labor input growth that were 3.5 percent and 2.5 percent slower, respectively. 33 No matter how you calculate it, Hong Kong SAR had about 2 percent higher TFP growth.

VII. Conclusion

To summarize, this chapter has four main findings:

  • Hong Kong SAR is roughly as competitive as the average OECD economy. Different measures tell slightly different stories. P-AC margins are generally higher in Hong Kong SAR, but P-MC margins are slightly lower. On almost all measures, Hong Kong SAR is neither the most nor the least competitive country in the 15-country sample.

  • Hong Kong SAR has become slightly less competitive in the last decade. The shift in industry mix to services has increased economy-wide average P-MC margins and value-added P-AC margins. Within industries, gross output P-AC margins have increased slightly in the last 10 years, as have value added P-AC margins in manufacturing (Figure 3), while value-added P-AC margins have fallen in service industries.

  • Margins are significantly more pro-cyclical in service industries compared to manufacturing. Since services are both the largest sector and have been rapidly increasing in importance in the last decade, overall margins should have also been becoming more pro-cyclical in Hong Kong SAR. Changes in margin cyclicality therefore would not be helpful in explaining any recent slowing of price adjustment.

  • Imperfect competition can seriously distort traditional growth accounting exercises, biasing TFP growth measures downward, especially in rapidly growing economies. Correcting the labor shares requires not just estimating market power, but also understanding factor fixity. An alternative approach would be to estimate the production function directly using micro data. Neither of these was feasible given the data available for this project, but with more data, this could be a fruitful area for further study.

Figure 3.
Figure 3.

Hong Kong SAR: P-AC Margins in Manufacturing and Service Sectors, 1986-97

Citation: IMF Working Papers 2000, 142; 10.5089/9781451856248.001.A001

References

  • Barro, Robert J., 1998, “Notes on Growth Accounting,” NBER Working Paper No. 654 (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

  • Bils, Mark, 1987, “The Cyclical Behavior of Marginal Cost and Price,” American Economic Review, Vol. 77 (December), pp. 838–55.

  • Bound, John, David A. Jaeger, and Regina M. Baker, 1995, “Problems with Instrumental Variables Estimation when the Correlation Between the Instruments and the Endogenous Explanatory Variable is Weak,” Journal of the American Statistical Association, 90, pp. 443–50.

    • Search Google Scholar
    • Export Citation

  • Dodsworth, John and Dubravko Mihaljek, 1997, Hong Kong SAR, China: Growth, Structural Change, and Economic Stability During the Transition, IMF Occasional Paper No. 52 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

  • Domowitz Ian, R. Glenn Hubbard, and Bruce C. Petersen, 1986, “Business cycles and the relationship between concentration and price-cost margins,” Rand Journal of Economics, 17, pp. 1–17.

    • Search Google Scholar
    • Export Citation

  • Domowitz Ian, R. Glenn Hubbard, and Bruce C. Petersen, 1988, “Market Structure and Cyclical Fluctuations in U.S. Manufacturing,” Review of Economics and Statistics, 70, pp. 55–66.

    • Search Google Scholar
    • Export Citation

  • Ellison, Glenn, 1994, “Theories of Cartel Stability and the Joint Executive Committee,” Rand Journal of Economics, 25, pp. 37–57.

  • Green, Edward J. and Robert H. Porter, 1984, “Non-cooperative collusion under imperfect price information,” Econometrica, 52, pp. 87–100.

    • Search Google Scholar
    • Export Citation

  • Hall, Robert E., 1988, “The Relation Between Price and Marginal Cost in U.S. Industry,” Journal of Political Economy, 96, pp. 921–47.

    • Search Google Scholar
    • Export Citation

  • Hall, Robert E., 1990, “Invariance properties of Solow’s productivity residual,” in Growth/Productivity/unemployment: Essays to Celebrate Bob Solow’s Birthday, ed. by P. Diamond (Cambridge, Massachusetts: MIT Press).

    • Search Google Scholar
    • Export Citation

  • Harrison, Ann E., 1994, “Productivity, imperfect competition, and trade reform,” Journal of International Economics, 36, pp. 53–73.

    • Search Google Scholar
    • Export Citation

  • Hausman, Jerry A., 1978, “Specification Tests in Econometrics,” Econometrica, 46, pp. 1251-1272.

  • Hsieh, Chang-Tai, 1997, “Measuring Technical Change,” mimeo.

  • Morrison, Catherine, 1990, “Market Power, Economic Profitability, and Productivity Growth Measurement: An Integrated Structural Approach,” NBER Working Paper No. 355 (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

  • Nelson, Richard and Howard Pack, 1995, “The East Asian Growth Miracle and Modern Growth Theory,” mimeo.

  • Oliveira Martins, Joaquim, Stefano Scarpetta, and Dirk Pilat, 1996, “Mark-up Ratios in Manufacturing Industries: Estimates for 14 OECD Countries,” OECD Economics Department Working Paper No. 62 (Paris).

    • Search Google Scholar
    • Export Citation

  • Rodrik, Dani, 1997, “TFPG Controversies, Institutions, and Economic Performance in East Asia,” NBER Working Paper No. 514 (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

  • Roeger, Werner, 1995, “Can Imperfect Competition Explain the Difference Between Primal and Dual Productivity Measures? Estimates for U.S. Manufacturing,” Journal of Political Economy, 103, pp. 316–330.

    • Search Google Scholar
    • Export Citation

  • Rotemberg, Julio J., and Garth Saloner, 1986, “A Supergame-theoretic Model of Price Wars During Booms,” American Economic Review, 76, pp. 390–407.

    • Search Google Scholar
    • Export Citation

  • Rotemberg, Julio J., and Michael Woodford, 1991, “Markups and the Business Cycle,” NBER Macroeconomics Annual 1991 (Cambridge, Massachusetts: MIT Press).

    • Search Google Scholar
    • Export Citation

  • Rotemberg, Julio J., and Michael Woodford, 1999 “The Cyclical Behavior of Prices and Costs,” NBER Working Paper No. 609 (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

  • Sarel, Michael, 1997, “Growth and Productivity in ASEAN Countries,” IMF Working Paper 97/97 (Washington: International Monetary Fund).

    • Search Google Scholar
    • Export Citation

  • Solow, Robert M., 1957, “Technical Change and the Aggregate Production Function,” Review of Economics and Statistics, 39, pp. 312–320.

    • Search Google Scholar
    • Export Citation

  • Young, Alwyn, 1995, “The Tyranny of Numbers: Confronting the Statistical Realities of the East Asian Growth Experience,” Quarterly Journal of Economics, 110, pp. 641-680.

    • Search Google Scholar
    • Export Citation

  • Young, Alwyn, 1998, “Paasche vs. Laspeyers: The Elasticity of Substitution and Bias in Measures of TFP Growth,” NBER Working Paper No. 663 (Cambridge, Massachusetts: National Bureau of Economic Research).

    • Search Google Scholar
    • Export Citation

  • Zitzewitz, Eric, 1999, “Real Exchange Rates, Trade Competition, Financial Pressure, and Productivity Growth in U.S. Manufacturing, 1958–92” (draft, Cambridge, Massachusetts: Massachusetts Institute of Technology).

    • Search Google Scholar
    • Export Citation

APPENDIX I

Derivation of P-MC Measurement Methodologies

All three methodologies (DHP 1988, Hall 1988, and Roeger 1995) start with a profit maximizing firm with a constant returns Cobb-Douglas production function facing a constant elasticity demand curve:

Q=A⋅Lα⋅Kβ;α+β=1Q=BP−σ;PQ=B−1/σ⋅Q1−1/σ;1<σ<∞

where P, Q, L, K and price, quantity, and labor and capital inputs, respectively. There are no fixed or adjustment costs. Firms maximize profits given factor costs W and R.

π=PQ−WL−RK

The first order conditions yield the following relationships:

P/MC=P/MR=(1−1/σ)−1=μWL/PQ=a=α⋅(1−1/σ)=α/μ

where a is the observed labor share. The expression for the change in the Solow residual (dTFP) is:

dTFP=dq−a⋅dl−(1-a)⋅dk=(μ-1)⋅a⋅(dl−dk)+d(lnA)(A1)

where lower case letters indicate logarithms. This equation comes from using the production function to substitute for dq and then using α = a. μ to remove α. Adding as (dl − dk) to both sides yields:

dq−dk=μ⋅a⋅(dl−dk)+d(lnA)(A2)

which can be estimated to yield an estimate of μ, treating the change in productivity (ln A) as an error term. Since the change in the output-capital ratio is likely to be correlated with the change in productivity, the factors that affect industry demand (B) are used as instrumental variables since they should be correlated with d(l − k) but not with productivity. This is the Hall (1988) methodology. It relies on the assumptions of constant returns to scale, the absence of labor fixity or adjustment costs, and the orthogonality of the demand instruments to productivity but not to the labor-capital ratio.34

The DHP (1988) method is closely related. The estimating equation is obtained by substituting the expression for a.(dl − dk) implied by (A2) into (A1). This yields:

dTFP=(μ-1)/μ⋅(dq−dk)+(1/μ)⋅d(lnA)=γ⋅(dq−dk)+(1−γ)⋅d(lnA)(A3)

where γ = (P − MC)/P = (μ -1)/μ, i.e. the P-MC margin. Like the Hall specification, this equation can be estimated using an demand instrument which affects the output-capital ratio but not productivity. An advantage of the DHP specification over the Hall specification is that demand instruments are likely to be more strongly correlated with changes in the output-capital ratio than with changes in the labor-capital ratio, giving an IV estimation using the DHP methodology more power.

Roeger (1995) takes a different approach, eliminating the productivity term by taking the difference of the primal (quantity-based) and dual (price-based) Solow residuals:

dTFP-dTFP-D=(dq+dp)−a⋅(dl+dw)−(1-a)⋅(dk+dr)dTFP-D=−dp−a⋅dw−(1-a)⋅dr(A4)

With a constant markup, the change in price will reflect the change in factor prices and the growth in productivity.

dp=α⋅dw+(1−α)⋅dr−d(lnA)

Substituting for α yields

dTFP−D=(μ−1)⋅a⋅(dr−dw)+d(lnA)

which implies that

dTFP-dTFP-D=(μ-1)⋅a⋅[(dl-dw)−(dk+dr)](A5)

Equations (A4) and (A5) can be manipulated in the spirit of the DHP methodology to yield

(dq+dp)-a⋅(dl+dw)−(1-a)⋅(dk+dr)=(μ-1)/μ⋅[(dq+dp)−(dk+dr)](A6)

This yields a direct estimate for the P-MC margin γ = (P − MC)/P = (μ - l)/μ. Roeger argues that since this equation omits the productivity term, it can be estimated via OLS without using demand instruments. Hausman (1978) tests in Table 4 support this assertion for U.S. data, but not for OECD data, suggesting that at least in the OECD data, one of the assumptions made by Roeger (e.g., no adjustment costs, constant returns to scale) is violated.

ANNEX I

Table 1.

Results for Detailed Hong Kong SAR Industries

article image