Value relevance of accounting information remains one of the most important areas in the financial accounting (Chalmers et al. 2010). Particularly, researches have concerned for the loss of value relevance of accounting information raising the speculation that accounting information may not been optimally used for the investors’ decision making. As a consequence, the relationship between accounting information and equity price tends to decline overtime. The results were ambiguous however. For example, while Collins et al. (2009), Alexander et al. (2012) and Francis et al. (2003) have provided some mixed proofs that value relevance is likely to increase overtime, other studies (Dontoh et al. 2004; Balachandran and Mohanram, 2011; Core et al. 2003) have provided robust findings that value relevance tends to decline over time in the long run.

We believe however, there are two determinants that may lead to ambiguous trends of value relevance. First, IFRS is continuously claimed to entail higher quality rather than Generally Accepted Accounting Principles (GAAP). As markets may make favorable anticipation for the convergence towards IFRS, the reliance of accounting information will be greater for firms adopting IFRS rather than their local GAAP counterparts. Second, different levels of accounting conservatism may affect the extent of accounting value relevance (Watts 2003). Most of the empirical evidences however, were conducted on the developed market. In contrary with the developed countries, firms in the emerging market entails different situations and environmental condition that may increase the relevance of accounting information. The technology and information in those markets may not be as sophisticated as what in the developed market. As a result, the market participants may solely rely on the use of reliable sources of financial accounting information (Dontoh et al. 2004).

Indonesia is a perfect research setting in order to test whether the IFRS plays an important role on the relationship between accounting conservatism and value relevance. The convergence process since 2009 until full convergence in 2012 can be used to figure out the trend of value relevance. Indonesian PSAK (Indonesian Statement of Financial Accounting Standards – Pernyataan Standar Akuntansi Keuangan) that has recently converged to International Financial Reporting Standards (IFRS) is continuously claimed to be more qualified, higher quality and as a consequence, to be more useful for decision making as compared to the Generally Accounting Accepted Principles (GAAP). Empirical evidences however, were stand divided whether the IFRS or its domestic GAAP counterpart, that may actually induce the value relevance of accounting information. For example, Barth et al. (2012, 2008), Schiebel (2006), Bartov et al. (2005), Palea (2014) and Liu and Liu (2007) showed that value relevance of book value and earnings are higher under IFRS as opposed to the GAAP. Interestingly, another extreme of the empirical findings (Hung and Subramanyam, 2007 and Daske et al. 2007 for instance) have provided some convincing evidence that capital market benefits are stronger in favor of local GAAP.

There are indeed some rationale to explain these conflicting findings. Ball (2006) and Daske et al. (2007) maintained that the accounting numbers of IFRS adoption would be more value relevant for countries that have weak enforcements of full disclosure to the their stakeholders rather than for countries that already have a strong foundation of governance systems, although Andréet al. (2015) proved that accounting conservatism declined after the IFRS adoption. We do not however, make such comparison as some authors have questioned the comparability issue due to interdependence of accounting standards and the underlying country specific factors and different economic, political and environment factors (Daske et al. 2008, Wherfritz and Haller 2014, Suadiye 2012, Barth et al. 2012, Soderstrom and Sun, 2007) .Instead, in contrast with the suggestion of Gjerde et al. (2008), the research setting in Indonesia will provide a comprehensive view of the value relevance of accounting information under the old (Indonesian GAAP) and the new regime (IFRS-based standards). This study test the hypothesis of incremental value relevance and its competing hypothesis of the superiority of one standard towards another (Dichev 2007, Ball 2006).

Another stream of research that may explain differences in the value relevance is the accounting conservatism (Balachandran and Mohanram, 2011; Ali and Hwang 2000 Holthausen & Watts, 2001; Grambovas, Giner, & Christodoulou, 2006 Bushman & Piotroski, 2006). Although it may be appealingto speculate that accounting conservatism should be blamed for the cause of the decline in the value relevance, the empirical evidences wereagain, not conclusive and scarce. For instance, although Balachandran and Mohanram (2011) maintained that value relevance of accounting information tends to be higher in the conservative as opposed to the non-conservative environment, they insisted that the decline of accounting conservatism was not affected by the conservatism per se. Interestingly, Watts (2003) found that increasing measurement errors in the conservative firms may always result in the decrease of reliability. Indeed, by recalling the trade-off between reliability vs relevance, we can speculate firmly that accounting conservatism increase value relevance of the firms. Ali and Hwang (2000) further stated that the increase of conservatism would lead to decrease of accounting conservatism, while Manganaris (2015) found the inverse relationship between conditional conservatism and value relevance after the IFRS adoption. Although without a robust multivariate quadratic regression analysis, Kousenidis et al. (2009) on the other hand, found the non-linear relationship between accounting conservatism and value relevance.

This research contributes to the accounting literature in several flourishing ways. First, although the value relevance of accounting information has been widely researched for more than two decades, most researches failed to recognize the fact that accounting is multi- faceted science that should be looked in a much-broader picture. The accounting conservatism and accounting standard impacts cannot be looked at in a piecemeal manner (Balachandran & Mohanram, 2011, Hu et al. 2014). We provide a comprehensive analyses by examining whether both the conditional and unconditional conservatism and standard implementation increase the value relevance.

Second, based on literature review on this area, most of the research provide a lot of speculations why the declining value relevance of accounting information exist (kindly see the findings from Alali and Foote 2012, Jiang and Stark 2013, Palea 2014, Wehfritz and Haller 2014). This research provides conclusive evidencesin Indonesia that test whether IFRS convergence have the significant impacts to enhance the value relevance of accounting information. Lastly, there is no empirical evidences specifically conducted in Indonesia that assess whether accounting information is still relevant and how the roles of accounting standards convergence and accounting conservatism affect the value relevance.

Indonesia is one of the largest archipelagos in the world that comprises of more than 17,000 islands and as of 2014, it is the home for more than 250 million people. With its exotic and rich resources, Indonesia was colonized for more than four decades where the Dutch was being the longest. It is not surprising therefore to find that during the early years after its independence, Indonesian laws, acts and regulations were heavily influenced by the Dutch systems, including its accounting systems.

However, not until 1954, the history of accounting profession in Indonesia (namely Indonesian Institute of Accountant or IAI) started to develop that was evident by enactment of Accountant Designation Act. In 1957, the first professional, independent accounting regulatory body was formed. The political reforms in 1967 was regarded as a cornerstone to move the directions, from Dutch-based into US-based accounting systems. Soon afterwards, the first US-based codified accounting systems, namely Indonesian Accounting Principles, was launched in 1973 although with an 8 years lag relative to US GAAP (Kusuma 2005). Although the influence of US GAAP remained strong until 2008, the plantoharmonize national accounting standards with the International Accounting Standards (IAS) has emerged since 1994. During this time, IAI has formally supported “the harmonization program initiated by International Accounting Standard Committee” (Maradona and Chand, 2014, 24).

Chand and Patel (2011) argued that IFRS implementation can be distinguished into five typologies: a) full adoption, b) selective adoption, c) country-specific modified IFRS adoption, d) maintaining old, national accounting standards but in line with IFRS, and e) totally maintaining national accounting standards. The convergence process in Indonesia falls in (b) and (c) in which the IFRS is gradually adapted into national accounting standards (Maradona and Chand 2014). The process however, could take years to fully complete.

With this regard, the convergence process of IFRS to national accounting standards have been started since 2007, that most of the new standards were fully adopted from IFRS. Until 2011, there were 35 accounting standards and 20 interpretations of financial accounting standards were endorsed by Indonesian Financial Accounting Standard Board with various effective dates (Maradona and Chand 2014). For example, although 1 standard has an effective date of 1 January 2009, the effective dates of 16 and 18 standards were on 1 January 2011 and 1 January 2012, respectively. Interestingly, IAI allowed for the early adoptions for 10 standards issued between 2009 and 2010. However, the time frame is very limited that observing the dates between the early adoptions and effective dates are difficult. As of 2012, the convergence process results Indonesian accounting standards were equivalent to the IFRS of 2009. IAI has publicly proclaimed that 2012 was the final phase of the convergence process.

Value relevance of accounting information can be simply identified if the accounting information is beneficial for investors’ decision making. In other words, accounting information is said to be relevant if it has a strong effect to market values (Barth et al. 2001). With this definition, value relevance can be measured using several proxies. First, value relevance can be operationalized using the adjusted R2 of regressing the book value and earnings to the stock price (Palea 2014,Lin and Chen 2005, Francis and Schipper 1999, Srinivashan and Narasimhan, 2012, Turel 2009, Alali and Foote, 2012, among others) as what is known as the price model of Feltham and Ohlson (1995) (hereafter, F-O model). Value relevance can also be measured using the adjusted R2of regressing the returns with the earnings level and earnings change, as known as the return model of Easton and Harris (1991) (hereafter, E-H model). Francis and Schipper (1999) on the other hand, provided some rationale for the value relevance in which the returns should have anticipated all available accounting information. All things considered, value relevance is the extent to which accounting information may be relevant for investment decision making and thus be impounded in stock prices.

The value relevance of accounting information can be traced back since 1960s through the seminal works of Ball and Brown (1968) and Beaver (1968). Since then, the topic received wider audience that tried to elaborate how the value relevance and information content may persist and under what condition it sustain. Researches on whether accounting information has lost its relevance however, have failed in providing consistent findings. Collins et al. (2009), Francis et al. (2003) among others have shown an increase in the value relevance of earnings over time, but Dontoh et al. (2004), Core et al. (2003) could not confirm such increase. Ebaid (2012) blamed for the higher reliance on aggregate earnings for explaining market values rather than earnings components.Although it can be speculated that as one of the emerging market, value relevance of accounting information may increase in Indonesia, empirical evidences were scarce. Therefore, we test the following null hypothesis:

H₀₁: The value relevance of accounting information remains constant over time

Conservatism can be simply characterized as a careful recognition of unforeseen contingencies in the future and an assuring mechanism that all risks and uncertainties have been proportionally considered in the financial statements. International Accounting Standards have concurred that accounting conservatism can be achieved when firms “make required predictions with uncertain factors and add certain degree of prudence in order not to raise assets or earnings nor to depress debt or expenses”. Simply stated, firms are encouraged to state the worse financial results when multiple alternatives in accountings methods and policies are available. As a consequence, the net assets and earnings will be likely to be understated (Kieso et al. 2004; Revsine et al. 2005). Beaver and Ryan (2000) identified the conservatism for any downward bias in book values relative to market values or as what Penman and Zhang (2002) and Xu et al. (2012) said as the unfavorable bias towards the losses rather than gains.

History of conservatism can be traced back before the implementation of accounting standards took place. Basu (1997) estimates that the philosophy of conservatism in the accounting practice can be traced back for at least 500 years ago, while Hu et al. (2014) also insists that conservatism had become one of the most important principles of earnings measurement and recognition before the existence of any accounting standards in the whole world. Similarly, Watts (2003) also maintained that accounting conservatism came into being from the needs of accountability of the contracts between the principal and agents and their rational interest distributions (Xu et al. 2012).

This study focused on two major classes of conservatism measures and relate them to the value relevance, which are the conditional and unconditional conservatism. Conditional conservatism, also sometimes known as the news dependent conservatism or asymmetric timeliness measure, is a measure of conservatism first pioneered by Basu (1997). He particularly maintained that a higher degree of verification is placed for good news in favor of bad news. As a result, the bad news is recognized faster than the bad news. On the other hand, while Basu (1997) focused on how the internal mechanism of conservatism takes place, we also looked at the end result of conservatism, also known as the unconditional conservatism. Unconditional conservatism can be defined as the prudence accounting reporting that is not influenced by economic reality (Balachandran and Mohanram 2011). This includes immediate, direct expensing of advertising and R& D expenses that could lead to faster omission of economic assets from balance sheet. Givoly and Hayn (2000) used the accruals in order to figure out whether the conservatism took place. They predicted that when the firms accelerate the losses and delay the gains, the accruals will be negatively accumulated. Theoretically speaking, the relationship between the price, book value and earnings should decline in the presence of high conservative accounting.

Researches on the relationship between accounting conservatism and value relevance however, have not yielded consistent findings. For example, Lev and Zarowin (1999) found that value relevance declines more for firms with the highest increase of Research and Development expenses. In contrary, Francis and Schipper (1999) found that value relevance does not decrease more for firms in the high technology rather than firms in the low technology industries. Although accounting conservatism studied were different between Lev and Zarowin (1999) and Francis and Schipper (1999) (while the earlier tested the level of conservatism while the later examined the change in conservatism), results were comparable. Watts (2003) on the other hand insisted that higher conservatism strengthened the relationship between earnings and price. It is apparent that the inconclusive findings were due to the different methods, measures and firms’ characteristics.

Motivated by the inconclusive findings, Kousenidis et al. (2009) found the non-linear relationship between conservative reporting and value relevance. It is interesting though to scrutinize the findings of Kousenidis (2009) in which the inconclusive findings were due to the non-linear nature of the relationship between accounting conservatism to the value relevance. We test their prediction in the sensitivity and robustness test in the next part of the study. Since the relationships are mixed and this area of research is still under detailed investigation, this study tested the following null hypothesis:

H₀₂: Accounting conservatisms do not affect the value relevance of accounting information

During the past decade, one of the most important research domain in the financial accounting area are the market effects of IFRS adoption and comparison of accounting information qualities (Hail et al., 2010a, 2010b). However, fewer evidence was presence regarding the comparability of accounting information as a result of IFRS vis a vis US GAAP. Studies in the financial accounting literature indicates that firms that applying the IFRS were likely to have more earnings quality rather than firms that adopting the domestic standards (Dimitropoulus et al. 2013, Barth et al. 2008, 2012). Leuz et al. (2003) and Ball et al. (2000) found that properties of accounting information, including accounting quality would be altered when there are differences in enforcement, incentives and application of accounting standards. Interestingly, Leuz and Verecchia (2000) and Leuz et al. (2003) cannot find conclusive evidence that differences in price volatility, trading volume and bid/ask spreads are influenced by the application of IFRS versus US GAAP in German. Bartov et al. (2005) found that US GAAP possess highest earnings response coefficients, while the IFRS and German Accounting Standards have the lower and lowest earnings response coefficient, respectively.

In another settings, Iatridis and Rouvolis (2010) reported that IFRS adoption in Greece increase relationship between the aggregate earnings and equity price. Francis and Shipper (1999) observed that the value relevance of earnings tend to decrease before the IFRS adoption took place, although no clear evidence whether the pattern continued afterwards nor changed to increase. In comparing the accounting quality between IFRS and US GAAP on terms of the aggregate earnings, Hughes and Sander (2008) found that US GAAP was of higher quality compared to IFRS, although the earnings from both standards are comparable. Although some of the empirical evidences presented above are from developed market, but they can be used to speculate that use of either GAAP or IFRS will increase the relevance of accounting information in the emerging markets (Barlev and Haddad 2007).

H₀₃: The value relevance of accounting information is not higher after the firms adopt high quality, international accounting standards (IFRS)

The previous hypotheses deals with the queries of the value relevance of accounting information that is partially affected by the accounting conservatism and IFRS. We however, also test whether the effect of accounting conservatism on value relevance may differ in the pre vs post IFRS implementation. Van der Meulen et al. (2007) that investigated the differences in the value relevance before and after IFRS adoption failed to provide some convincing evidences whether IFRS adoption outperformed the US GAAP on terms of the value relevance. On the other hand, decrease of value relevance after the IFRS implementation has also been observed by Khanaga (2011), although in certain cases, although he found strong evidence of incremental information content of cash flow during post IFRS adoption.

As conservatism may also play some important roles in the value relevance of accounting information, it is interesting to also figure out whether the effect of accounting conservatism to the value relevance will change after the adaptation of IFRS-based standards. The interaction product of accounting conservatism and IFRS implementation is intended to fill the missing gaps in the accounting literature about this area. Nevertheless, no research had been conducted to shed some light regarding the roles of IFRS implementation on accounting conservatism and value relevance relations, and thus this study estimates the following null hypothesis:

H₀₄: the accounting conservatism does not influence the value relevance of accounting information even after the implementation of high quality accounting standards (IFRS)

Current study collected data from industrial firms in Indonesia during 2003 - 2014. Although some accounting standards permit the early adoption of IFRS-based PSAK, we simply differentiate the pre IFRS period as the years from 2003 – 2012 and post IFRS period as the years of 2013 and 2014. Data were collected from Bureau van Dijk, Osiris that initially consisted of 429 publicly listed industrial firms with a balanced sample of 5148 firm year observations.

Furthermore, as the conditional conservative of Basu measures of the timely recognition of bad vs good news requires 5 yearly data for robust estimation, we remove some firms that did not publish their financial reports for 5 consecutive years. As a consequence, the final pooled, unbalanced cross-sectional sample includes 244 firms with 2166 firm year observations. We also removed the missing values observed in the data and thus yields different number of firm year observations for the uni-variate and/multivariate analyses.We also use local currency units (1 US $ = IDR 13,500 as per 1 July 2015) instead of standardized, US Dollar measures to maintain the comparability of the variables among the firms. Therefore, we able to control for the currency exchange deviations bias.

Conditional conservatism can be simply characterized as a prudent recognition of unforeseen contingencies in the future and an assuring mechanism that all risks and uncertainties have been proportionally considered in the financial statements. In order to measure conditional conservatism, current study uses the measurement of Basu (1997) that could assess the asymmetrical speed of bad news compared to the good news recognition. He modeled the conservatism as:

$\frac{E P S_{i, t}}{P_{i, t-1}}=\beta_0+\beta_1 D R_{i, t}+\beta_2 R_{i, t}+\beta_3 R . D R_{i, t}+\varepsilon_{i, t}$ (1a)

where EPS_i,t is the earnings per share of firm i at year t; DR_i,t is a dummy variable, where 1 indicates the presence of bad news that is simply proxied by negative returns of firm i at year t, and 0 otherwise.; R_i,t is annual raw return of firm i at year t; P_i,t is the closing price of firm i at year t; and R.DR_i,t is the interaction product of dummy variable (DR) and stock returns (R) of firm i at year t. In this case the interaction product takes the value of R_i,t if the returns are negative and zero otherwise. As been clearly pointed out by Basu (1997), an asymmetric recognition of bad against good news will present in the data if the interaction product of 𝛽₃ will be significantly positive. Moreover, while 𝛽₁ indicates how the firms recognize the good news 𝛽₁ + 𝛽₃ reflects earnings recognition of bad news. The intercept of 𝛽₀ may be used as a proxy of the firms’ average cost of capital (Pope and Walker 1999).

We create three sets of conditional conservatism as suggested by Balachandran and Mohanram (2011), as also been used by Kousenidis (2009) using two steps. First, we ranked the value of 𝛽₃, ascendingly ordered. Second, we grouped the sampled firms into high conservatism (HI-CON), medium conservatism (MED-CON) and low conservatism (LOW- CON). The high and low conservatism consisted of the highest and lowest 30% of 𝛽₃, while the medium conservatism consisted of the remaining 40%. This grouping is the basis for most of the analyses pertaining to the conditional conservatism. We also use the raw value of 𝛽₃ as a robustness test regarding conditional conservatism.

Conservatism can also be measured using the news-independent or ex ante conservatism, that is also known as unconditional conservatism (Beaver and Ryan 2000). In contrast with conditional conservatism, the recognition of accounting expenses/news is conducted prior to the occurrence of economic news. Unconditional Conservatism in this study employs the negative accrual first pioneered by Givoly and Hayn (2000). The main lemma of the negative accrual (NEGACC) is that firms tend to utilize the flexibility in the accruals to postpone a recognition of gains and to expedite a recognition of losses (Yunos et al. 2012). Since they found that non-operating accruals are the main source of flexibility that cannot be nullified by any increase in the operating accruals, the Negative Accruals is operationalized as:

NEGACC ={(NI- depreciation )-(Δ INV+Δ DEBTORS+Δ OCA -Δ CREDITORS-Δ OCL*-1 (1b)

where NI represents the net income and depreciation is the net depreciation expenses, Δ INV is the change in the net inventory, Δ DEBTORS represents the change in the debtors fund, Δ OCA is the change in other current assets, Δ CREDITORS denotes for the change in creditors fund and Δ OCL represents the change in other current liabilities. We then multiplied NEGACC with -1 so that the higher value NEGACC indicates the higher unconditional conservatism.

Value relevance can be defined as a condition in which earnings are relevance for the investor’s decision making. The only way to figure out whether the accounting information (i.e. earnings) is value relevance if the strong relation can be observed between earnings and price. There are many methods that can be used to measure the value relevance, however we only use two of most widely used ones, namely: the price model of Feltham and Ohlson (1995) and the return model of Easton and Harris (1991). The price model of F-O regressed stock price on earnings and book value:

$P_{i, t}=\gamma_0+\gamma_1 B V P S_{i, t}+\gamma_2 E P S_{i, t}+\varepsilon_{i, t}$ (2)

where P_i,t is the share price of firm i at year t; BVPS_i,t is the per share book value of equity of firm i at year t and EPS_i,t is the earnings per share of firm i at year t . On the other hand, the return model of Easton and Harris (1991) can be expressed as:

$R E T_{i, t}=\alpha_0+\alpha_1 \frac{E P S_{i, t}}{P_{i, t-1}}+\alpha_2 \frac{\Delta E P S_{i, t}}{P_{i, t-1}}+\varepsilon_{i, t}$ (3)

Where RET_i,t is the annual return of firm i a time t, ΔEPS_i,t is the difference of EPS of firm i at the beginning and at the end of period t.

There are two sources of information in which the value relevance of information can be gathered: 1) the significance relations of earnings per share and book value per share for the price model and earnings level and earnings change to returns for the return models (Barth 1994, Venkatachalam 1996, Hassan and Mohd-Saleh, 2010 among others), or 2) from the fit measure (Adjusted R2) (Balachandran and Mohanram 2011, Collins et al. 1997 among others). The earlier is usually used to test the value relevance of a particular accounting information (i.e. EPS, BVPS, earnings level, earnings change) while the R2 measures the value relevance for the set of accounting information in a particular equation (Filip and Raffournier, 2010). A value of adjusted R2 was also commonly used to identify the trend of value relevance over time.

We also address the possible issue of the econometrics regarding the functions in Eq. 2 and Eq. 3. As been highlighted by Klein and Marquardt (2006), information of negative earnings tends to be less relevance than positive earnings. This asymmetric timeliness in the value relevance function may lead to the under-estimated bias of goodness of fit. Therefore, we also control for the loss in the price and return functions of value relevance:

𝑃_i,𝑡 = 𝛾₀ + 𝛾₁𝐵𝑉𝑃𝑆_i,𝑡 + 𝛾₂𝐸𝑃𝑆_i,𝑡 + 𝛾₃𝐿𝑂𝑆𝑆_i,𝑡 + 𝛾₄𝐿𝑂𝑆𝑆_i,𝑡 ∗ 𝐸𝑃𝑆_i,𝑡 + 𝛾₅𝐿𝑂𝑆𝑆_i,𝑡 ∗ 𝐵𝑉𝑃𝑆_i,𝑡+ε_i,𝑡 (2a)

𝑅𝐸𝑇_i,𝑡 = 𝛼₀ + 𝛼₁ $\frac{E P S_{i, t}}{P_{i, t-1}}$ + 𝛼₂ $\frac{\Delta E P S_{i, t}}{P_{i, t-1}}$ + 𝛼₃𝐿𝑂𝑆𝑆_i,𝑡 + 𝛼₄𝐿𝑂𝑆𝑆_i,𝑡 ∗ $\frac{E P S_{i, t}}{P_{i, t-1}}$ + 𝛼₅𝐿𝑂𝑆𝑆_i,𝑡 ∗ $\frac{\Delta E P S_{i, t}}{P_{i, t-1}} \varepsilon_{i, t}$ (3a)

where LOSS is a dummy variable that takes a value of 1 if a firm i has negative earnings in year t and 0 other wise. Other variables are as defined above. Furthermore, inter-industry heterogeneity may also lead to the value relevance, and thus should be controlled:

𝑃_i,𝑡 = 𝛾₀ + 𝛾₁𝐵𝑉𝑃𝑆_i,𝑡 + 𝛾₂𝐸𝑃𝑆_i,𝑡 + 𝛾₃𝐿𝑂𝑆𝑆_i,𝑡 + 𝛾₄𝐿𝑂𝑆𝑆_i,𝑡 ∗ 𝐸𝑃𝑆_i,𝑡 + 𝛾₅𝐿𝑂𝑆𝑆_i,𝑡 ∗ 𝐵𝑉𝑃𝑆+ 𝛾₆𝐼𝑁𝐷_i,𝑡 + 𝛾₇𝐼𝑁𝐷_i,𝑡 ∗ 𝐸𝑃𝑆 + 𝛾₈𝐼𝑁𝐷_i,𝑡 ∗ 𝐵𝑉𝑃𝑆_i,𝑡 + ε_i,𝑡 (2b)

𝑅𝐸𝑇_i,𝑡 = 𝛼₀ + 𝛼₁ $\frac{E P S_{i, t}}{P_{i, t-1}}$ + 𝛼₂ $\frac{\Delta E P S_{i, t}}{P_{i, t-1}}$ + 𝛼₃𝐿𝑂𝑆𝑆_i,𝑡 + 𝛼₄𝐿𝑂𝑆𝑆_i,𝑡 ∗ $\frac{E P S_{i, t}}{P_{i, t-1}}$ + 𝛼₅𝐿𝑂𝑆𝑆_i,𝑡 ∗ $\frac{\Delta E P S_{i, t}}{P_{i, t-1}}$ + 𝛼₆𝐼𝑁𝐷_i,𝑡 + 𝛼₇𝐼𝑁𝐷_i,𝑡 ∗ $\frac{E P S_{i, t}}{P_{i, t-1}}$ + 𝛼₈𝐼𝑁𝐷_i,𝑡 ∗ $\frac{\Delta E P S_{i, t}}{P_{i, t-1}}$ + ε_i,𝑡 (3b)

Where in line with McIver (2014) IND is simply raw code of global industry classification standard (GICS) derived from eight digit level reflecting their industrial groups (excluding the financial). Other variables are as previously defined.

The value relevance is measured using the adjusted R2 of all the models. In order to test for the decline or increase of value relevance, we run the time-series regrssion, again, based on three models: a) base model (eq. 2 and 3), b) loss control model (eq. 2a and 3a), and c) loss and industry control model (eq. 2b and 3b).

Where NEGACC is the unconditional conservatism as defined from equation (1b), IFRS is a dummy variable of 1 for the firms mandatory adoption of IFRS-based standard since 2012, and 0 is otherwise. Other variables are previously defined. Equation 5a and 6a test the impact of unconditional conservatism on the value relevance using the price and return model, respectively. Equations 5b and 6b, on the other hand examines whether the effects of unconditional conservatism on value relevance differs upon the pre vs post IFRS adoption for the price and return models, respectively. While equations 5c and 6c scrutinizes the effect of loss vs profit making on the relationship between unconditional conservatism on value relevance for the price and return models, respectively.

Furthermore, we also follow the suggestions from Brown et al. (1999) and Balachandran and Mohanram (2011) that as the change in the value relevance may be mechanically driven by the change in the coefficient of variation of price and book value per share, and thus it should be controlled. However, our results showed that these has a very small change in the value relevance, and hence unreported here. The trends of value relevance are tested as followings:

VALREL_p.2n =ά0+ά1 YEAR_i,t+ε (4a)
VALREL_r.2n =ά0+ά1 YEAR_i,t+ε (4b)

VALREL_p.2n is the adjusted R² of the value relevance measure for the price model in equation (2), (2a) or (2b), while VALREL_r.2n is the adjusted R² of the value relevance measure for the return model in equation (3), (3a) or (3b). YEAR is a discrete variable, simply reflecting the year in which the value relevance is measured (results were displayed in the last three rows of table 2). If the ά1 in those equations are positive and significant, the increased value relevance can observed

Another measure of conservatism that we employ in this case is the unconditional conservatism of negative accrual that is estimated in the value relevance equations. We test the following price regression models of F-O to test the impact of unconditional conservatism on value relevance:

_i,𝑡 = 𝛾₀ + 𝛾₁𝐵𝑉𝑃𝑆_i,𝑡 + 𝛾₂𝐸𝑃𝑆_i,𝑡 + 𝛾₃𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 + 𝛾₄𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 ∗ 𝐸𝑃𝑆_i,𝑡 + 𝛾₅𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 ∗ 𝐵𝑉𝑃𝑆_i,𝑡 + e (5a)

𝑃_i,𝑡 = 𝛾₀ + 𝛾₁𝐵𝑉𝑃𝑆_i,𝑡 + 𝛾₂𝐸𝑃𝑆_i,𝑡 + 𝛾₃𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 + 𝛾₄𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 ∗ 𝐸𝑃𝑆_i,𝑡 + 𝛾₅𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 ∗ 𝐵𝑉𝑃𝑆_i,𝑡 + 𝛾₆𝐼𝐹𝑅𝑆_i,𝑡 + 𝛾₇𝐼𝐹𝑅𝑆_i,𝑡 ∗ 𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 ∗ 𝐸𝑃𝑆_i,𝑡 + 𝛾₈𝐼𝐹𝑅𝑆_i,𝑡 ∗ 𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 ∗ 𝐵𝑉𝑃𝑆_i,𝑡 + e (5b)

𝑃_i,𝑡 = 𝛾₀ + 𝛾₁𝐵𝑉𝑃𝑆_i,𝑡 + 𝛾₂𝐸𝑃𝑆_i,𝑡 + 𝛾₃𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 + 𝛾₄𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 ∗ 𝐸𝑃𝑆_i,𝑡 + 𝛾₅𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 ∗ 𝐵𝑉𝑃𝑆_i,𝑡 + 𝛾₆𝐼𝐹𝑅𝑆_i,𝑡 + 𝛾₇𝐼𝐹𝑅𝑆_i,𝑡 ∗ 𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 ∗ 𝐸𝑃𝑆_i,𝑡 + 𝛾₈𝐼𝐹𝑅𝑆_i,𝑡 ∗ 𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 ∗ 𝐵𝑉𝑃𝑆_i,𝑡 + 𝛾₉𝐿𝑂𝑆𝑆_i,𝑡 + 𝛾₁₀𝐿𝑂𝑆𝑆_i,𝑡 ∗ 𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 + 𝛾₁₁𝐿𝑂𝑆𝑆_i,𝑡 ∗ 𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 ∗ 𝐸𝑃𝑆_i,𝑡 + 𝛾₁₂𝐿𝑂𝑆𝑆_i,𝑡 ∗ 𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 ∗ 𝐵𝑉𝑃𝑆_i,𝑡 + e (5c)

While the return models of E-H are as follows:

𝑅𝐸𝑇_i,𝑡 = 𝛼₀ + 𝛼₁ $\frac{E P S_{i, t}}{P_{i, t-1}}$ + 𝛼₂ $\frac{\Delta E P S_{i, t}}{P_{i, t-1}}$ + 𝛼₃𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 + 𝛼₄𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 ∗ $\frac{E P S_{i, t}}{P_{i, t-1}}$ + 𝛼₅𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 ∗ $\frac{\Delta E P S_{i, t}}{P_{i, t-1}}$ + e (6a)

𝑅𝐸𝑇_i,𝑡 = 𝛼₀ + 𝛼₁ $\frac{E P S_{i, t}}{P_{i, t-1}}$ + 𝛼₂ $\frac{\Delta E P S_{i, t}}{P_{i, t-1}}$ + 𝛼₃𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 + 𝛼₄𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 ∗ $\frac{E P S_{i, t}}{P_{i, t-1}}$ + 𝛼₅𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 ∗ $\frac{\Delta E P S_{i, t}}{P_{i, t-1}}$+ 𝛼₆𝐼𝐹𝑅𝑆_i,𝑡 + 𝛼₇𝐼𝐹𝑅𝑆_i,𝑡 ∗ 𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 ∗ $\frac{E P S_{i, t}}{P_{i, t-1}}$ + 𝛼₈𝐼𝐹𝑅𝑆_i,𝑡 ∗ 𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 ∗ $\frac{\Delta E P S_{i, t}}{P_{i, t-1}}$ + e (6b)

𝑅𝐸𝑇_i,𝑡 = 𝛼₀ + 𝛼₁ $\frac{E P S_{i, t}}{P_{i, t-1}}$ + 𝛼₂ $\frac{\Delta E P S_{i, t}}{P_{i, t-1}}$ + 𝛼₃𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 + 𝛼₄𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 ∗ $\frac{E P S_{i, t}}{P_{i, t-1}}$ + 𝛼₅𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 ∗ $\frac{\Delta E P S_{i, t}}{P_{i, t-1}}$+ 𝛼₆𝐼𝐹𝑅𝑆_i,𝑡 + 𝛼₇𝐼𝐹𝑅𝑆_i,𝑡 ∗ 𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 ∗ $\frac{E P S_{i, t}}{P_{i, t-1}}$ + 𝛼₈𝐼𝐹𝑅𝑆_i,𝑡 ∗ 𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 ∗ $\frac{\Delta E P S_{i, t}}{P_{i, t-1}}$ + 𝛼₉𝐿𝑂𝑆𝑆_i,𝑡 + 𝛼₁₀𝐿𝑂𝑆𝑆_i,𝑡 ∗ 𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 + 𝛼₁₁𝐿𝑂𝑆𝑆_i,𝑡 ∗ 𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 ∗ $\frac{E P S_{i, t}}{P_{i, t-1}}$ + 𝛼₁₂𝐿𝑂𝑆𝑆_i,𝑡 ∗ 𝑁𝐸𝐺𝐴𝐶𝐶_i,𝑡 ∗ $\frac{\Delta E P S_{i, t}}{P_{i, t-1}}$ + e (6c)

Table 1 presents the Pearson correlations and descriptive statistics of the variables. While most of the important variables are statistically significant, our results indicate no serious multi-collinearity issues. The correlational statistics also hinted the value relevance of earnings and book value. More specifically, the strongest value relevance appeared in the medium conservatism group for both the price and return models.

Our results of trends in the value relevance arevisualized in Figure 1 and Table 2. As expected, using the similar methods of Collins et al. (1997), Balachandran and Mohanram (2011), Filip and Raffournier (2010), among others, we found the consistent findings of the accelerating value relevance of accounting information in the Indonesian stock market. However, the two models that we used, namely the price model of Feltham and Ohlson (1995), and the return model of Easton and Harris(1991), did not provide the conclusive findings. While the price model of F-O indicates the strong incremental value relevance, E-H did not, even after controlling for the variances of losses and industrial groups. It is likely that the presence of structural break in the pooled year of 2003 -2007 may be the reason for the rejection of incremental value relevance, as clearly displayed in Figure 1 panel B.

The Adj R² of the price model in of 2003-2004 is 63.61%, 65.85%, and 67.77% for the base, loss control and loss and industry control models respectively (see Table 2 panel A). For the panel model of 2003-2008, the Adj. R² increased to 72.53%, 73.39%, and 76.3% and significantly increased again for the year of 2003 – 2014 yielding the Adj R² of 81.99%, 82.32%, and 83.15% for the base, loss control and loss and industry controlled models respectively. When we regress the value relevance index of E-H (adj. R²) on the year (Table 2, panel B), we cannot find any significant relationship. The estimated parameter for the base, loss and industrial controlled models are -0.326 (t-stat= -0.879), -0.594 (t-stat = - 0.982) and -0.097 (t-stat = -1.466), respectively. The variations of the value relevance that are explained by the yearly variance is very small, ranging from 7.9% to 10.3%.

We also further scrutinize whether the incremental value relevance may be different depending upon the levels of accounting policy. As displayed in Table 3, we found the statistically difference among the value relevance of the firms with the high, medium and low levels of conservatism (F-Stat = 49.957, p < 0.001). In this regards, the medium level of conservatism (MEAN = 94.79%) have the highest value relevance as compared to the low (MEAN = 79.80%) and high of conservatism groups (MEAN = 51.49%). Bonferroni tests of mean difference also indicate the statistically difference of value relevance among the groups of high vs medium, high vs low and medium vs low conservatisms (due to simplicity, results are not shown).

Interestingly, after we made the groupings for the level of conservatism, the declining value relevance is present among the firms with the medium accounting conservatism group. The yearly decline of value relevance is -0.4% (Table 3 panel B). While the value relevance in the 2003 -2004 is 98.44%, smooth declination can be observed in the subsequent periods (value relevance of 2003 – 2014 is 93.8%). This decline however, cannot be confirmed for the value relevance of the firms with the high (Table 3 panel A) or low accounting conservatism (Table 3 panel C). For example, although the accounting conservatism is quite high for the high conservatism group (mean = 51.4%) and low (mean = 79.8%), our results do not indicate the decline or incremental value relevance. The t-values from regressing value relevance on year are 0.236 and -0.102 for the high and low group of accounting conservatism, respectively.These results are in part concur with Balachandran and Mohanram (2011) maintaining that steady, moderate accounting conservatism tend to have a larger decline in accounting conservatism.

We found that both EPS and BVPS are significantly value relevant for most of the analyzed models. However, while we documented the positive coefficient of EPS, the BVPS coefficient was negative. This phenomenon is commonly found in some researches (e.g. see Chalmers et al. 2010). While it is important to figure out the value relevance of EPS and BVPS per se, this is not the focus of current study. As in hypothesis 2, this study deals with impact of IFRS implementation to the value relevance. From Table 4 (panel A), it can be found that the interaction product of IFRS*EPS is positively significant (γ4A = 8.951; t = 27.4) but not for the BVPS measure (γ5A = 0.0709, t = 0.89). It indicates that the value relevance of earnings is higher post IFRS period while IFRS does not change the value relevance of the book value.

After we control for the effect of negative earnings, we can also figure out that the value relevance of earnings is lower for loss making firms (γ7A = -9.212, t = -5.06), while the value relevance of book value is higher for loss making firms (γ8A = 0.194; t = 3.15). We can also conclude that there is intra-industry difference of value relevance (γ10A = 0.18, t = 8.11 and γ11A = 0.020, t = 3.32; for the EPS and BVPS measures respectively). Furthermore, we also found thatthe higher value relevance of book value in the post IFRS implementation is more apparent for firms with negative earnings (γ14A = 0.87; 3.594).

When we split the samples into three portfolios based on various levels of conservatism, the results are somehow more interesting. More specifically, since IFRS is measured as a dummy variable, this study can figure out whether the value relevance is higher or lower in the post vis a vis prior IFRS convergence. In the high conservatism group (Table 4 panel B), the value relevance of earnings is higher in the post IFRS period (γ4B = 3.685, t = 6.091), but not for the value relevance of book value (γ5B = -0.117, -1.131). Both the value relevance of earnings (γ7B = -1.953, t = -0.999) and book value (γ8B = 0.0423, t = 1.101) however do not differ for loss vs profit making firms. The value relevance of earnings differences can be observed among diverse industrial groups (γ10B = -0.292, t = -7.985). The value relevance of loss making firms were constant during the pre and post IFRS implementation periods. Our results also found that among the medium level of conservatism group in the post IFRS implementation, the value relevance of earnings (book value) is higher (γ4/C = 11.157, t = 20.999), while the value relevance of book value is lower (γ5/C = -0.810, t = -5.233). Moreover, the value relevance of earnings is lower for loss making firms (γ7/C = -14.595, t = -2.817), while the value relevance of book value is higher for loss making firms (γ8/C = 2.052, t = 4.759). Although the value relevance of both earnings and book value for the loss making firms do not differ for the pre vs post IFRS (γ13.C = -16.059; t = -0.019 and γ14.C = 0.837; t = 0.938 for the earnings and book value measures respectively), different industry classifications affects the earnings and book value relevancies.

For the low level of conservatism groups (Table 4 panel D), it can be observed that both value relevance of earnings (γ4.D = 4.774; t = 7.036) and book value (γ5.D = 0.524; t = 3.720) are higher after IFRS implementation. However, the value relevance of earnings is lower among the loss making firms (γ7.D = -12.865; t = -4.100), but higher for the value relevance of book value (γ8D = 0.519, t = 1.873). Different value relevance is strongly apparent among diverse industries (γ10.D = 0.078; t = 1.751 and γ11.D = 0.098; t = 9.046 for the earnings and book value measures, respectively). The value relevance measures during pre vs post IFRS do not differ for the loss vs profit making firms.