A. Issuance Status

Government bonds in Korea consist of four major categories: Treasury Bonds (KTBs), Foreign Exchange Stabilization Fund Bonds (FESFBs), National Housing Bonds (NHBs), and Treasury Stabilization Bonds (TSBs). Among these, KTBs are the primary instrument issued to finance fiscal operations and serve as the benchmark in Korea’s fixed-income markets. KTBs are issued in multiple maturities—including 2-, 3-, 5-, 10-, 20-, 30-, and 50-year fixed-rate bonds—along with 10-year inflation-linked bonds.

Figure 1 shows the monthly outstanding balance and issuance volume of KTBs from 2015 to 2021. During this period, both measures increased steadily, with a particularly sharp rise in 2020–2021 due to the large-scale fiscal stimulus following the COVID-19 pandemic. Between 2015 and 2019, the annual increase in outstanding balances was between KRW 20 trillion and KRW 47 trillion. However, in 2020 and 2021, the increase surged to KRW 115.3 trillion and KRW 116.9 trillion, respectively, pushing the outstanding balance to KRW 726.8 trillion in 2020 and KRW 843.7 trillion in 2021.

FIGURE 1.

MONTHLY OUTSTANDING BALANCE AND ISSUANCE OF TREASURY BONDS FROM 2015 TO 2021

Source: Ministry of Economy and Finance, Treasury Bond Market Website.

For the purposes of this study, what matters is the timing and magnitude of these supply shifts, as they provide meaningful variation for identifying the yield effects of issuance shocks. Accordingly, the revised manuscript limits the discussion to major issuance changes and moves detailed historical tables and additional descriptive statistics to the Appendix.

Figure 2 summarizes the evolution of issuance composition by maturity. Two structural features stand out. First, the Korean government expanded the maturity spectrum over time, introducing a 50-year bond in 2016 and reintroducing the 2-year KTB in 2021. Second, the share of long-term issuance increased markedly: the 30-year KTB, for example, rose from 11.1% of annual issuance in 2015 to 26.5% in 2021. This shift reflects stronger demand from long-horizon institutional investors, particularly insurers, and the government's efforts to extend its debt maturity profile. As a natural counterpart, the share of short-term issuance declined over the same period.

FIGURE 2.

ISSUANCE COMPOSITION OF TREASURY BONDS BY MATURITY FROM 2015 TO 2021

Source: Ministry of Economy and Finance, Treasury Bond Market Website.

The increase in long-term issuance, supported in part by foreign investors who hold a significant share of medium- and long-term KTBs, contributed to a gradual extension of the average maturity to 11.66 years by 2021. This maturity-lengthening strategy has reduced refinancing risk and supported more stable debt management.

B. Treasury Bond Buyback (Buy-back) Operations

This section examines the recent status of Treasury bond buybacks (early redemptions) in Korea. A buyback refers to the repurchase and cancellation of Treasury bonds by the government before their maturity. Buybacks can be classified into two types based on the funding source.

First, a conventional buyback is conducted by issuing new Treasury bonds to finance the repurchase of existing ones (refinancing). As this approach simply replaces one bond with another, it does not alter the total outstanding balance of Treasury bonds.

Second, buybacks can also be financed using surplus fiscal resources such as excess tax revenues. In this case, the outstanding amount of Treasury bonds is reduced, constituting a net redemption.

The primary objective of buyback operations is to smooth the maturity profile of outstanding government debt. If a large volume of Treasury bonds matures within a short period, refinancing could require sudden large-scale issuances, potentially distorting the market and causing sharp interest rate hikes. By dispersing bond maturities through buybacks, the government can mitigate rollover risks and prevent sudden spikes in borrowing costs.

In October of 2020, the Korea government introduced an emergency buyback system designed to strengthen the resilience of the government bond market and facilitate rapid policy responses to external and internal shocks. In practice, emergency buybacks were implemented in August and November of 2021 to counteract increased volatility and rapid interest rate hikes amid monetary policy normalization (see cases ④ and ⑤ below).

The scale of buyback operations between 2015 and 2021is shown in Figure 3. The Ministry of Economy and Finance (MOEF) pre-announces the schedule and amount of buybacks through its monthly KTB issuance plans. However, unexpected buybacks or unanticipated expansions in buyback volumes also occur. In this study, six such “unexpected buyback” cases are identified and summarized, as follows:

TABLE 1

SUMMARY OF EMERGENCY BUYBACK OPERATIONS

Source: MOEF Press Release; Monthly Treasury Bond Issuance Plan; BOK Financial Market Commentary; BOK Financial Market Key Indicators.

FIGURE 3.

SCALE OF TREASURY BONDS BUYBACKS FROM 2015 TO 2021

Source: Ministry of Economy and Finance, “Treasury Bonds 2021,” 2022.

1. On or around October 25, 2017, the buyback volume was unexpectedly increased from the originally planned KRW 1 trillion to KRW 1.6 trillion. According to the MOEF’s “October Treasury Bond Issuance Plan” press release dated September 21, three buybacks were scheduled: KRW 1.5 trillion on September 27, KRW 1 trillion on October 18, and KRW 1 trillion on October 25. While the latter was pre-announced, the actual operation on October 25 involved a KRW 1.6 trillion buyback, reflecting an unexpected increase.

2. On or around November 15, 2017, a scheduled KRW 1 trillion buyback was canceled. In January of 2019, the MOEF explained that this decision was unavoidable, taking into account discussions on additional deficit bond issuances, market conditions, and fiscal liquidity at the year-end.

3. On December 12, 2018, a buyback worth KRW 2.7 trillion was conducted. However, the official "December 2018 Treasury Bond Issuance Plan" had announced a KRW 2 trillion operation on that date. Thus, the buyback amount was unexpectedly expanded by KRW 0.7 trillion.

4. On August 25, 2021, the first-ever emergency buyback in Korea history was implemented. Unlike conventional buybacks, emergency buybacks are announced at the time of operation without prior notice, in response to market conditions. Anticipating the Bank of Korea's first rate hike during the Monetary Policy Committee (MPC) meeting on August 26, the government conducted KRW 1 trillion emergency buybacks on August 25 and August 31. These operations were unexpectedly announced on August 20. Notably, unlike typical maturity-smoothing buybacks, these constituted net redemptions and represented the largest redemption effort backed by supplementary budgets. The operations contributed to a reduction in the national debt-to-GDP ratio and mitigated bond market volatility.

5. On November 5, 2021, facing rapid interest rate increases due to rising concerns about global and domestic policy normalization, the government executed an additional KRW 2 trillion emergency buyback. Following a sharp rise in KTB yields in October, the government convened an emergency meeting on November 2 and announced the buyback, which was implemented immediately on November 5. The move signaled the government’s commitment to market stabilization, resulting in a reversal of the rate spike and a return to relative stability in the bond market. In total, emergency buybacks in 2021 amounted to KRW 4 trillion (cases ④ and ⑤), helping ease volatility and suppress yield spikes.

6. On December 28, 2021, a KRW 700 billion buyback was carried out to smooth maturity distributions using the remaining limit for market stabilization operations. This buyback plan was announced at the Treasury Bond Issuance Strategy Council meeting on December 22. According to the Bank of Korea’s financial market commentary, yields on 3-year KTBs fell on December 27 in anticipation of the buyback.

A. The Necessity of Controlling for Bond Demand (Capital Supply) Variables

In this study, we estimate the impact of Treasury bond issuance on government bond yields using both a linear regression analysis and the macro-finance dynamic Nelson–Siegel (MF-DNS) model developed by Diebold et al. (2006). The MF-DNS model allows us to examine how exogenous changes in macroeconomic variables influence the entire yield curve. The following section outlines the MF-DNS framework and the details of our empirical specification.

To analyze the effect of bond issuance on yields accurately, it is important to account for the supply of funds in the government bond market. This can be illustrated using a simple supply–demand framework. Figure 4 presents the supply and demand curves for funds in the Treasury bond market, where the x-axis denotes the quantity of funds and the y-axis shows the Treasury yield. The supply of funds is upward-sloping because higher yields incentivize market participants to provide more capital. In contrast, the government’s demand for funds—represented by bond issuance—is determined exogenously by fiscal needs and is therefore depicted as a vertical line.

An increase in issuance corresponds to a rightward shift of the demand curve, as shown in Figure 4(a). When fiscal spending rises and the government issues more bonds, the market equilibrium moves from point A to point B, and yields increase from level A to level B, assuming that the supply curve remains unchanged. However, if the supply of funds rises at the same time—for example due to greater market liquidity or a stronger investor appetite—the supply curve will also shift to the right, as illustrated in Figure 4(b). In this case, the equilibrium moves from point B to point C and yields decline from level B to level C. This demonstrates that even when bond issuance expands, the observed yield response may be small or even negligible if the supply of funds adjusts simultaneously. Therefore, identifying the pure effect of Treasury issuance requires controlling for shifts in the supply of funds.

FIGURE 4.

GOVERNMENT BOND DEMAND/SUPPLY CURVE

To avoid overstating the role of demand controls, we clarify that our use of primary dealer (PD) bidding data is not intended to identify bond demand shocks in an exogenous or structural sense. We fully acknowledge that, within the MF-DNS framework, isolating pure demand shocks or fully controlling for all dimensions of investor demand is inherently difficult—a limitation shared by most empirical term-structure studies. Our aim is not to assert that PD bid amounts provide a comprehensive measure of aggregate demand. Rather, we use these data as a practical, high-frequency proxy that captures part of the short-term variation in the market absorption capacity at the time of issuance.

Accordingly, the PD bid data should be understood as a behavioral indicator of revealed willingness-to-purchase in the primary market, reflecting dealers’ liquidity positions, inventory conditions, and portfolio preferences. We do not assume that these bids trace out a structural downward-sloping demand curve, nor do we view them as fully exogenous demand shifters. Instead, they help mitigate potential omitted-variable bias by controlling for observable auction-level fluctuations in buying interest that coincide with supply shocks. This is especially relevant in segmented or partially segmented markets, where primary-market bidding behavior may evolve in ways not fully captured by secondary-market yields or macro-financial variables.

Finally, we explicitly acknowledge that PDs also participate actively in the secondary market, and their bidding decisions may incorporate expectations about future issuance events or broader market conditions. For this reason, we refrain from characterizing PD bids as “controlling for demand” in a strict structural sense. Rather, we position them as a complementary control that enhances the empirical identification of issuance effects relative to specifications without any auction-based demand information. This revised exposition reflects the methodological limitations highlighted by the referee and clarifies the realistic scope of what the PD bidding data can—and cannot—capture within our analytical framework.

B. Model Specification: The MF-DNS Model

To estimate how changes in Treasury bond issuance and buybacks affect the entire KTB yield curve, this study employs a macro-finance dynamic Nelson–Siegel (MF-DNS) model. The MF-DNS framework is particularly suitable because it links movements in the yield curve to macroeconomic conditions while preserving the flexibility and empirical tractability of the original Nelson–Siegel representation. Importantly, the model provides a structure within which the influence of issuance shocks can be separated from contemporaneous demand factors—an identification challenge that simple regression approaches cannot fully overcome.

A crucial innovation of this study is the incorporation of variables that directly address this identification problem. While bond yields are shaped not only by supply shocks but also by shifts in investor demand, standard yield curve models do not capture short-run demand fluctuations well. To address this, we augment the MF-DNS framework with two sets of variables. First, we include macroeconomic factors that broadly influence the demand for government bonds, in this case monetary policy, inflation conditions, and real activity indicators. Second, we incorporate auction-level bidding amounts from primary dealers (PDs and PPDs), which provide a high-frequency, behavior-based proxy for the market absorption capacity at the time of issuance. These bidding data help capture demand-side pressures that cannot be explained by macro variables alone—allowing us to better isolate the supply-driven component of yield movements. This choice of variables is not merely an extension of the DNS structure but is a central element of our identification strategy.

We also allow the influence of issuance shocks to differ before and after the COVID-19 pandemic by including a COVID dummy variable, reflecting the structural shift in both issuance patterns and investor behavior during this period. Furthermore, we distinguish issuance volumes by maturity bucket (short-, medium-, and long-term maturities) to examine how supply shocks propagate differently across the yield curve—an analysis that is essential for understanding the maturity-dependent effects of the bond supply.

While the full DNS and MF-DNS frameworks can be expressed as a state-space system with latent level, slope, and curvature factors, we keep the technical exposition concise and relegate full equations and derivations to the Appendix. The essential point is that each latent factor evolves according to macroeconomic conditions and issuance-related variables, allowing us to trace out how supply shocks affect yields of all maturities.

Because the augmented MF-DNS model involves a large number of parameters and potentially strong multicollinearity between macro variables and latent factors, we adopt a Bayesian estimation strategy, which is widely used in the literature for high-dimensional term-structure models. Bayesian estimation improves numerical stability and reduces the risk of converging to local optima. Posterior distributions of parameters and latent factors are obtained via Markov chain Monte Carlo (MCMC) sampling, with computational details presented in the Appendix. The main text focuses on the conceptual motivation behind the approach, consistent with the referee’s guidance to avoid unnecessary technical elaboration.

Through this integrated specification—combining macroeconomic controls, auction-based demand proxies, issuance-by-maturity variables, and a regime indicator for the COVID period—the MF-DNS model provides a coherent empirical framework that yields cleanly identified estimates of how issuance and buyback shocks shape the Korean government bond yield curve.

C. The Model Framework: MF-DNS

Building on Kang, Lee, and Kang (2021), we employ an MF-DNS model to estimate how KTB issuance affects the yield curve. While Kang, Lee, and Kang (2021) focused on the net issuance of selected maturities (10- and 20-year bonds), we extend their approach by incorporating total issuance across all maturities, further disaggregated into short-term (2–3 year), medium-term (5–10 year), and long-term (20–30 year) categories. This allows us to examine maturity-specific supply effects along the curve.

We also include a COVID-19 dummy variable to capture potential structural changes in the relationship between issuance and yields before and after the pandemic.

Before presenting the macro-augmented model, we briefly outline the baseline dynamic Nelson–Siegel (DNS) framework, which represents the yield curve using three latent factors—level, slope, and curvature—in a parsimonious and empirically tractable form.

where

In the macro-augmented MF-DNS model, each factor l_t , s_t , and c_t is allowed to depend on its own lag and a vector of contemporaneous macroeconomic variables, m_t . The transition equation is defined as follows:

Combining the measurement and transition equations, the model can be expressed in state-space form as follows:

Measurement equation: y_t =Λf_t + e_t

Transition equation: f_t = μ + G_fff_t-1 + G_fmm_t + ν_t

Here, f_t = (l_t, s_t, c_t)' and e_t and ν_t are mutually independent.

From these, we can derive the effect of macroeconomic variables m_t on the contemporaneous yield curve y_t .

In line with standard practice in the Nelson–Siegel and MF-DNS literature, we fix the decay parameter λ at 0.0609, following the estimates reported in Diebold and Li (2006) and subsequent empirical applications for Korea. Fixing λ ensures comparability with prior studies and prevents potential identification issues that arise when λ is estimated jointly with the latent factors, especially in shorter samples.

A. Data Description

We use Korea Treasury bond yield data from January of 2015 to June of 2021. While some earlier studies begin with 2001 based on available yield data, we restrict our sample to 2015 onward because bid data from primary dealers (PDs and PPDs)—crucial for controlling for bond demand during auctions—are available only from 2015.

Yield data are based on auction yield rates (awarded yields) provided by the Ministry of Economy and Finance (MOEF). Because awarded yields represent the coupon rate determined during the auction, they reflect the government’s effective borrowing cost. In contrast, spot yields (zero-coupon rates) reflect closing prices in the secondary market and may be influenced by non-issuance factors. Therefore, this study uses awarded yields to isolate the effect of the issuance volume on borrowing costs. We use five maturities: 3-, 5-, 10-, 20-, and 30-year KTBs.

Additional macroeconomic variables include Korea’s policy rate, the U.S. federal funds rate, the 10-year U.S. Treasury yield, the KRW/USD exchange rate, the Consumer Price Index (CPI), the Industrial Production Index (IPI), and the Treasury turnover ratio.

For the analysis of the impact of buybacks on yields, we use daily data, as buyback-induced yield changes occur over very short time frames and would not be captured by monthly frequency data.

B. Descriptive Statistics

Figure 5 illustrates the trends in Treasury bond (KTB) yields across different maturities from January of 2015 to June of 2021. At any given point in time, yields tend to be higher for longer-maturity KTBs. This upward-sloping yield curve reflects the fact that government bond yields are composed of the expected path of short-term interest rates and a term premium. The term premium compensates investors for holding longer-duration bonds, which are exposed to greater risks. Thus, unless the expected short-term rate is projected to decline sharply in the future, yields typically increase with maturity.

FIGURE 5.

TREND OF TREASURY BOND ISSUANCE RATES BY MATURITY AND THE BANK OF KOREA BASE RATE

Source: Internal data from the Ministry of Economy and Finance; Economic Statistics System, Bank of Korea.

This characteristic is further confirmed by the summary statistics presented in Table 2. During the sample period, the average yields by maturity were as follows: approximately 1.58% for 3-year bonds, 1.77% for 5-year bonds, 2.02% for 10-year bonds, 2.07% for 20-year bonds, and 2.10% for 30-year bonds, clearly demonstrating the positive term structure.

TABLE 2

DESCRIPTIVE STATISTICS FOR THE LINEAR REGRESSION ANALYSIS

Source: Internal data from the Ministry of Economy and Finance; Economic Statistics System, Bank of Korea.

The time-series pattern of KTB yields generally mirrors the movements in the Bank of Korea’s policy rate, as shown in Figure 6. However, following the surge in bond issuance in 2020, yields began to rise despite the continued low level stated in the policy rate. Korea was in a monetary easing cycle from July of 2012 until November of 2017. Typically, changes in KTB yields precede changes in the policy rate. Accordingly, KTB yields declined gradually from 2015 through mid-2016, anticipating policy rate cuts. As expectations for monetary tightening emerged, yields gradually increased until early 2018. Subsequently, the policy rate entered another easing cycle from November of 2018, and the outbreak of COVID-19 pushed interest rates further to historic lows.

FIGURE 6.

BIDDING AMOUNT AND BIDDING RATIO TRENDS BY MATURITY

Source: Author’s calculations based on internal data from the Ministry of Economy and Finance.

In addition to the bond issuance volume, this study considers other factors that may influence KTB yields. One such factor is the turnover ratio of government bonds, which serves as a proxy for market liquidity. Figure 7 depicts the turnover ratios for different maturities from 2015 to June of 2021. The turnover ratio is calculated as follows:

In this study, the primary dealer (PD) bid data refer to the total amount of bids submitted by each PD in competitive KTB auctions, aggregated at the auction level. These bids contain the full schedule of price–quantity pairs that PDs are willing to purchase, thereby capturing their actual buying intentions at the time of issuance. Because PDs actively participate in both markets, i.e., primary and secondary, their auction bids reflect real-time information pertaining to inventory positions, liquidity conditions, and portfolio demand. As such, the bid amounts serve as a high-frequency, action-based measure of underlying demand conditions that is more sensitive to short-term market dynamics than broader indicators such as holdings or turnover ratios.

Against this backdrop, we use PD bid amounts as a control variable to partial out contemporaneous demand-side variation when estimating the pure supply effect of government bond issuance. While we do not claim that the PD bidding data offer a complete measure of bond demand, they meaningfully account for short-run fluctuations in the market absorption capacity and help reduce potential omitted-variable bias in the identification of supply shocks within the MF-DNS framework. In this sense, PD bids are best interpreted as a practical and behaviorally grounded proxy for demand conditions at the moment of issuance, rather than as a fully exogenous demand shifter.

Lastly, we explicitly acknowledge that PD bidding behavior may incorporate expectations about forthcoming issuance schedules, meaning that bid data should not be viewed as entirely exogenous. Our intention is not to suggest otherwise. Instead, we position bid data as a complementary control that refines identification by removing observable demand-side movements that coincide with issuance events. This clarification aligns the empirical strategy given the data limitations and highlights the incremental value of incorporating auction-level microdata relative to existing studies.

At the same time, we acknowledge that PD bidding behavior may not be fully exogenous. In particular, PDs may adjust their bidding strategies in anticipation of heavy issuance periods, especially when the expected supply calendar places pressure on inventory management or the balance-sheet capacity. Such forward-looking adjustments imply that bid volumes reflect expectations about future issuance as well as contemporaneous demand, thereby introducing potential endogeneity into the control variable. For this reason, we do not interpret PD bids as a structural or exogenous demand shifter. Instead, we position them as a complementary control that improves empirical identification, while explicitly recognizing the limitations inherent in this approach.

Figure 6 provides critical evidence of auction-level demand conditions, showing the evolution of both bid amounts and bid-to-cover ratios across different KTB maturities. These two indicators summarize the intensity of primary dealer (PD and PPD) demand at each auction and therefore serve as direct, behavior-based measures of the market’s short-run absorption capacity at the moment of issuance. Unlike macroeconomic variables, which capture broad demand conditions at a lower frequency, these auction indicators reveal how strongly investors are willing to absorb the new government bond supply at each auction.

A key insight from Figure 6 is the sustained decline in bid-to-cover ratios since 2016, even as total issuance volumes increased markedly. While bid amounts have risen gradually, they have not kept pace with the rapid growth in supply, resulting in lower coverage ratios across most maturities. This divergence highlights a tightening of marginal demand in the primary market. When issuance expands faster than demand, the auction becomes more supply-constrained, making yields more sensitive to issuance shocks. This pattern is particularly evident in the medium- and long-term segments, where demand has historically been thinner and more volatile.

The trends documented in Figure 6 are therefore not merely descriptive; they motivate a core component of our identification strategy. By including bid amounts in the empirical model, we control for these auction-specific demand fluctuations that coincide with issuance shocks. Without such controls, estimated issuance effects would conflate supply shocks with shifts in the bidding pressure, especially during periods when bid-to-cover ratios are falling. Furthermore, because the decline in coverage ratios varies across maturities, Figure 6 also explains why issuance shocks may have heterogeneous effects along the yield curve—a result confirmed in our empirical findings.

A. The Impact of Government Bond Issuance on Market Interest Rates

The empirical analysis in this study is based on the dynamic Nelson-Siegel (DNS) model with macroeconomic factors. To estimate the effect of increased Treasury bond issuance on the yield curve more accurately, it is essential to account for the dynamic interactions across maturities and time. These include the historical influences of the level, slope, and curvature on current yields. The dynamic structure of the Nelson-Siegel framework allows for such comprehensive modeling of intertemporal and cross-maturity interactions.

In constructing the macroeconomic variables used in the model, the Industrial Production Index (IPI) and Consumer Price Index (CPI) are transformed into year-over-year growth rates. Variables with deterministic trends and integrated order one (I(1)) characteristics—such as the exchange rate—are converted to stationary form via log-differencing (i.e., first differences of the log-transformed values).

Table 3 presents the estimated effect of total KTB issuance on yields across maturities. The results indicate that increased issuance leads to higher yields across the entire yield curve. Specifically, a KRW 1 trillion increase in issuance is associated with an estimated rise in KTB yields of 2.50 to 2.86 basis points on average. All estimates are statistically significant.

TABLE 3

MF-DNS ANALYSIS RESULTS: TOTAL TREASURY BOND ISSUANCE—TREASURY BOND YIELDS BY MATURITY

Note: All estimated coefficients are expressed in percentage points (%p), and multiplying the values by 100 allows interpretation in basis points (bp).

Source: Author’s calculations based on internal data from the Ministry of Economy and Finance and data from the Economic Statistics System of the Bank of Korea.

These estimates are derived under a specification that controls for shifts in the supply curve, including bid amounts and relevant macroeconomic variables. In contrast, when bid amounts are not controlled for, the estimated impact of a KRW 1 trillion increase in issuance is only 0.5 to 1 basis point on average, and the estimates lack statistical significance as the maturity duration increases. For 5- and 10-year bonds, a KRW 1 trillion increase in issuance raises yields by approximately 2.64 basis points and 2.77 basis points, respectively—higher than the corresponding estimate of 2.50 basis points for 3-year bonds by 0.14bp and 0.27bp. For 20- and 30-year bonds, the estimated effects are 2.83bp and 2.86bp, respectively—approximately 0.06bp and 0.09bp higher than that for the 10-year bond.

This greater sensitivity of long-term bond yields can be attributed to two main factors: the rising share of long-term bond issuance and the relatively weak demand base for long-term KTBs. Since 2017, the proportion of long-term KTBs has steadily increased—from 21.1% in 2015 to 31.9% in 2021. As the share of long-term issuance expanded, the yield curve response to the total issuance volume became more concentrated at the long end.

A useful way to interpret the magnitude of our estimates is to consider how typical fluctuations in issuance translate into changes in market yields. As shown in Figure 1 and Table 2, the standard deviation of monthly KTB issuance during the sample period is approximately KRW 3.36 trillion—substantially larger than the KRW 1 trillion unit used for our coefficient estimates. Applying our baseline estimate of a 2.5–2.9 basis point yield increase per KRW 1 trillion issuance, a “typical” monthly fluctuation of roughly KRW 3 trillion would raise yields by about 7.5–8.7 basis points. This magnitude is economically meaningful in the Korean bond market, where yield changes of this size can influence funding conditions for households and firms, affect the pricing of long-term institutional portfolios, and alter the government’s fiscal borrowing costs.

Interpreting the estimates in this context highlights the policy relevance of issuance decisions: even routine monthly adjustments in issuance volumes can generate yield movements large enough to affect macro-financial conditions. This underscores the importance of managing the timing and scale of bond issuance—particularly during periods of elevated volatility or strained demand conditions—to mitigate unnecessary increases in the cost of capital across the economy.

Beyond changes in issuance proportions by maturity, the impact of increased issuance on yields may differ across short-, medium-, and long-term maturities even when controlling for issuance shares. To investigate this further, we conduct an additional estimation using the MF-DNS model, disaggregating issuance volumes into three categories: short-term (2- and 3-year), medium-term (5- and 10-year), and long-term (20-, 30-, and 50-year) bonds. This allows us to assess whether the effect of issuance on yields is systematically differentiated by maturity segment.

According to the results summarized in Table 4, when short-term issuance increases by KRW 1 trillion, yields on 3-year and 5-year KTBs rise by approximately 2.85 basis points (bp) and 2.34bp, respectively. In contrast, yields on medium- and long-term bonds increase by only 0.6–1.58bp, with estimates not statistically significant. When medium-term issuance increases by KRW 1 trillion, yields on 10-, 20-, and 30-year bonds rise by 2.65bp, 2.45bp, and 2.15bp, respectively, while short-term yields increase by only 1.10–1.76bp, also lacking statistical significance. Finally, a KRW 1 trillion increase in long-term issuance leads to increases of 2.62bp for 20-year and 3.15bp for 30-year bonds. In contrast, yields on short- and medium-term bonds do not exhibit statistically significant changes. These findings suggest that long-term issuance—and to some extent, medium-term issuance—affects long-term yields more significantly than short-term yields.

TABLE 4

MF-DNS ANALYSIS RESULTS: SHORT-, MEDIUM-, AND LONG-TERM TREASURY BOND ISSUANCE—TREASURY BOND YIELDS BY MATURITY

Note: All estimated coefficients are expressed in percentage points (%p), and multiplying the values by 100 allows interpretation in basis points (bp).

Source: Author’s calculations based on internal data from the Ministry of Economy and Finance and data from the Economic Statistics System of the Bank of Korea.

Taken together, the disproportionate impact on long-term yields appears to be driven by two factors: the growing share of long-term issuance in recent years, and the sensitivity of long-term yields to increases in both medium- and long-term issuance volumes.

Furthermore, estimation results from the model incorporating an interaction term between issuance volume and outstanding balance reveal that as the outstanding balance increases, the yield impact of new issuance decreases. Specifically, when the outstanding stock increases by KRW 100 trillion, the marginal effect of a KRW 1 trillion issuance on yields falls by approximately 0.25–0.37bp. Although the estimates lack strong statistical significance, they suggest that the expanded demand base for Korea Treasury Bonds in recent years may have attenuated the yield effects of new issuance.

Table 5 presents estimates of the effect of the total issuance volume on yields by maturity, again separating the pre- and post-COVID periods. These results indicate that the yield impact per KRW 1 trillion of issuance decreased after the onset of the pandemic. With bid amounts controlled for, yields increased by an average of 3.00–3.91bp per KRW 1 trillion before COVID-19, with all estimates statistically significant. After COVID-19, the yield impact fell to 2.07–2.46bp on average, with statistically significant estimates for all maturities except the 3-year bond.

TABLE 5

MF-DNS ANALYSIS RESULTS 2: TOTAL TREASURY BOND ISSUANCE—TREASURY BOND YIELDS BY MATURITY

Note: 1) Total issuance refers to the sum of issuance volumes of 3- to 30-year Treasury bonds; 2) The data analysis period is from January of 2015 to June of 2021; 3) The pre-COVID-19 period refers to January of 2015 to December of 2019, and the post-COVID-19 period refers to January of 2020 to June of 2021; 4) All estimated coefficients are expressed in percentage points (%p), and multiplying the values by 100 allows interpretation in basis points (bp).

Source: Author’s calculations based on internal data from the Ministry of Economy and Finance and data from the Economic Statistics System of the Bank of Korea.

These findings are consistent with the earlier results from the interaction model in Table 6, which also suggested that the marginal impact of issuance on yields has declined since 2020. This may reflect a broadening of the demand base for KTBs. In particular, increased confidence among foreign investors in Korea’s bond market may have played a key role. Foreign investors have continued to participate actively in the KTB market. According to the IMF, even during the 2013 ‘taper tantrum’, when the currencies of many emerging Asian economies depreciated sharply, Korea’s exchange rate and government bond yields remained stable. This stability may have led foreign investors to reclassify Korea as a “safe haven” (Hansen and Krogstrup, 2019). Such a perception has likely supported the expansion of demand for Korea Treasury Bonds in recent years, thereby reducing the magnitude of yield increases in response to new issuance.

Finally, Table 5 also confirms that in both pre- and post-COVID periods, the effect of issuance on yields increases with the bond maturity duration. Before COVID-19, yields on 5- and 10-year bonds rose by 3.35bp and 3.68bp per KRW 1 trillion upon issuance, compared to 3.00bp for 3-year bonds. Yields on 20- and 30-year bonds increased by 3.86bp and 3.91bp, respectively—0.18bp and 0.23bp higher than for the 10-year bond. All estimates are statistically significant.

After COVID-19, yields on 5- and 10-year bonds increased by 2.22bp and 2.36bp, respectively, relative to 2.07bp for the 3-year bond. For 20- and 30-year bonds, the estimated effects were 2.44bp and 2.46bp, respectively—0.08bp and 0.10bp greater than for the 10-year bond.

Additional evidence supports the interpretation that the weaker post-COVID issuance effect reflects a broadening of the demand base for Korean Treasury Bonds. As shown in Figure 7, foreign investors’ holdings of KTBs increased markedly after 2020. Both the level and the share of foreign holdings rose steadily, moving from below 15% prior to the pandemic to nearly 30% by 2023. This expansion in foreign participation coincides with a period in which overall outstanding government debt grew substantially, indicating that foreign investors absorbed a meaningful portion of the increased supply.

FIGURE 7.

FOREIGN INVESTORS IN KOREA TREASURY BONDS: HOLDINGS, SHARE, AND NET INVESTMENT

Source: Infomax.

The panel on the right in Figure 3 also shows that net foreign investment in KTBs became persistently positive during 2020–2022, with several months exhibiting inflows exceeding KRW 20–40 trillion. These inflows occurred despite heightened global volatility, suggesting that foreign investors increasingly viewed Korean government bonds as a comparatively safe and stable asset. This shift in investor perception is consistent with the broader narrative that Korea’s bond market has begun to function as a “safe haven” relative to other emerging markets.

Taken together, the rise in foreign holdings and strong net inflows imply that the demand base for KTBs expanded significantly in the post-COVID period. This structural strengthening of demand provides a concrete explanation of the empirical finding in Table 5 that the yield impact of issuance (per KRW 1 trillion) declined after 2020. With a larger share of issuance absorbed by foreign investors and other long-horizon institutions, the marginal pressure of the additional supply of bonds on yields naturally diminished.

B. The Impact of Treasury Bond Buybacks on Market Interest Rates

This section analyzes the impact of government bond buybacks, which—although opposite in direction to the interest rate effects of bond issuance—operate through the same underlying mechanism of altering the supply of government bonds. In Korea, buybacks are generally classified as either routine operations aimed at smoothing the maturity profile and emergency buybacks conducted for market stabilization. Routine buybacks are typically anticipated by market participants, making it unlikely that they will generate significant yield movements at the time of announcement or execution. In contrast, emergency buybacks are unanticipated and tend to produce more immediate and noticeable yield effects, often on the day of the announcement or even before the actual operation is carried out.

To clarify why buyback effects may differ from issuance effects, we provide a brief theoretical rationale. Under a standard benchmark, increases and decreases in the bond supply would yield symmetric price responses. However, market microstructure considerations suggest that buyback effects are likely to be asymmetric. Emergency buybacks are usually conducted during periods of market stress or reduced liquidity, when the marginal value of liquidity is heightened. In such circumstances, reducing the outstanding supply can help relieve dealers’ inventory imbalances, ease balance-sheet constraints, and compress liquidity premia more strongly than what would be predicted by a simple reversal of issuance shocks.

Furthermore, issuance and buyback operations interact differently with preferred-habitat or segmented-market structures. Large-scale issuance often exerts upward pressure on yields in maturity segments where investor demand is relatively inelastic, thereby raising term premia. Conversely, emergency buybacks are typically targeted at segments experiencing yield dislocations or pronounced supply–demand imbalances. Removing bonds from these stressed segments can therefore have a disproportionately large marginal effect. This provides a plausible theoretical basis underlying why buyback effects need not mirror issuance effects on a one-for-one basis.

Our empirical expectation of asymmetry thus does not rely on the assumption of different structural elasticities. Rather, it reflects the fact that issuance and buyback operations occur under systematically different market conditions and serve different economic functions. We explicitly incorporate this interpretation into the revised manuscript.

Because the effects of emergency buybacks tend to materialize over short time horizons, we rely on daily rather than monthly or quarterly data to capture yield responses accurately in these situations. Initial results using the dynamic Nelson–Siegel model—including both routine and emergency buybacks—are presented in Tables 6 and 7. These results show no clear patterns or statistically significant effects, likely due to the overwhelming dominance of routine buybacks that are already incorporated into market expectations. To address this, we separately analyze the subset of emergency buybacks.

TABLE 6

EFFECTS OF TOTAL BUYBACKS ON TREASURY YIELDS (EXCLUDING BUYBACK ANNOUNCEMENT DUMMIES)

Note: 1) Dummies were created for −1 day, +1 day, +2 days, and +3 days relative to the buyback announcement date; 2) The pre-COVID-19 period refers to January of 2015 to December of 2019, and the post-COVID-19 period refers to January of 2020 to June of 2021; 3) All estimated coefficients are expressed in percentage points (%p), and multiplying the values by 100 allows interpretation in basis points (bp).

Source: Author’s calculations based on internal data from the Ministry of Economy and Finance and data from the Economic Statistics System of the Bank of Korea.

TABLE 7

EFFECTS OF TOTAL BUYBACKS ON TREASURY YIELDS (INCLUDING BUYBACK ANNOUNCEMENT DUMMIES)

Note: 1) Dummies were created for −1 day, +1 day, +2 days, and +3 days relative to the buyback announcement date; 2) The pre-COVID-19 period refers to January of 2015 to December of 2019, and the post-COVID-19 period refers to January of 2020 to June of 2021; 3) All estimated coefficients are expressed in percentage points (%p), and multiplying the values by 100 allows interpretation in basis points (bp).

Source: Author’s calculations based on internal data from the Ministry of Economy and Finance and data from the Economic Statistics System of the Bank of Korea.

The results, reported in Table 8, indicate that emergency buybacks lead to statistically significant declines in government bond yields. Across the full sample period, a KRW 1 trillion emergency buyback lowers yields by approximately 1.91 basis points for 3-year bonds, 2.12 basis points for 5-year and 10-year bonds, 2.09 basis points for 20-year bonds, and 2.08 basis points for 30-year bonds. These results suggest that the yield-lowering effect is most pronounced for maturities of ten years or less. Although the 20- and 30-year yields fall slightly less than the 5- and 10-year yields, they still decline more than the 3-year yield by around 0.18 basis points.

TABLE 8

EFFECT OF EMERGENCY BUYBACKS ON TREASURY YIELDS (EXCLUDING BUYBACK ANNOUNCEMENT DUMMIES)

Note: 1) Dummies were created for −1 day, +1 day, +2 days, and +3 days relative to the buyback announcement date; 2) The pre-COVID-19 period refers to January of 2015 to December of 2019, and the post-COVID-19 period refers to January of 2020 to June of 2021; 3) All estimated coefficients are expressed in percentage points (%p), and multiplying the values by 100 allows interpretation in basis points (bp).

Source: Author’s calculations based on internal data from the Ministry of Economy and Finance and data from the Economic Statistics System of the Bank of Korea.

Importantly, the effects differ significantly between the pre- and post-COVID-19 periods. Before the pandemic, a KRW 1 trillion emergency buyback reduced yields by 0.72 basis points for 3-year bonds, 1.07 basis points for 5-year bonds, 1.89 basis points for 10-year bonds, 2.39 basis points for 20-year bonds, and 2.55 basis points for 30-year bonds—all statistically significant. These findings indicate that prior to COVID-19, buybacks exerted particularly strong effects on long-term maturities.

After the onset of COVID-19, however, the pattern shifts. Yields declined by 3.52 basis points for 3-year bonds, 3.54 basis points for 5-year bonds, 2.58 basis points for 10-year bonds, 1.95 basis points for 20-year bonds, and 1.74 basis points for 30-year bonds. The estimates are statistically significant for the 3- and 5-year maturities only, implying that the impact of emergency buybacks became larger in magnitude overall but more concentrated in short-term maturities.

This post-pandemic pattern likely reflects the combination of an accommodative monetary policy and sharply rising issuance, which together constrained the degree to which long-term yields could decline. Emergency buybacks during this period appear to have been timed to alleviate temporary supply–demand imbalances facing primary dealers (PDs and PPDs), thereby easing short-term market stress. The stronger yield response in the short-term segments may also reflect the fact that banks and securities firms—who are more sensitive to short-term liquidity pressures—hold a larger share of short-term bonds, whereas insurers, who are less liquidity-constrained, dominate long-term bond holdings. Consequently, the market impact of emergency buybacks was more pronounced at the short end of the yield curve.

TABLE 9

EFFECT OF EMERGENCY BUYBACKS ON TREASURY YIELDS (INCLUDING BUYBACK ANNOUNCEMENT DUMMIES)

Note: 1) Dummies were created for −1 day, +1 day, +2 days, and +3 days relative to the buyback announcement date; 2) The pre-COVID-19 period refers to January of 2015 to December of 2019, and the post-COVID-19 period refers to January of 2020 to June of 2021; 3) All estimated coefficients are expressed in percentage points (%p), and multiplying the values by 100 allows interpretation in basis points (bp).

Source: Author’s calculations based on internal data from the Ministry of Economy and Finance and data from the Economic Statistics System of the Bank of Korea.